Doreian, Batagelj, Ferligoj - Generalized Blockmodeling
-
Upload
larsrordam -
Category
Documents
-
view
222 -
download
0
Transcript of Doreian, Batagelj, Ferligoj - Generalized Blockmodeling
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 149
Photo Vladimir Batagelj UNI-LJ
Blockmodeling
Anuska FerligojUniversity of Ljubljana
Budapest 16 September 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 249
A Ferligoj Cluster Analysis 1
Outline1 Introduction 1
2 Cluster clustering blocks 2
4 Equivalences 4
9 Establishing blockmodels 9
11 Indirect approach 11
19 Direct approach 19
24 Generalized blockmodeling 24
26 Pre-specified blockmodeling 26
31 Symmetric-acyclic blockmodel 31
34 Blockmodeling of vauled networks 34
35 Blockmodeling of two-mode network 35
39 Blockmodeling of three-way networks 39
42 Application Clustering of Slovenian sociologists 42
48 Open problems in blockmodeling 48
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 349
A Ferligoj Cluster Analysis 1
Introduction
The goal of blockmodeling is to reduce a
large potentially incoherent network to a
smaller comprehensible structure that can be
interpreted more readily Blockmodeling
as an empirical procedure is based on the
idea that units in a network can be grouped
according to the extent to which they are
equivalent according to some meaningful
definition of equivalence (structural (Lorrain
and White 1971) regular (White and Re-
itz 1983) generalized (Doreian Batagelj
Ferligoj 2005)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 449
A Ferligoj Cluster Analysis 2
Cluster clustering blocksOne of the main procedural goals of blockmodeling is to identify in a given network
N = (U R) R sube U times U clusters (classes) of units that share structural charac-
teristics defined in terms of R The units within a cluster have the same or similar
connection patterns to other units They form a clustering C = C 1 C
2 C
k
which is a partition of the set U Each partition determines an equivalence relation
(and vice versa) Let us denote by sim the relation determined by partition C
A clustering C partitions also the relation R into blocks
R(C i C j) = R cap C i times C j
Each such block consists of units belonging to clusters C i and C j and all arcs leading
from cluster C i to cluster C j If i = j a block R(C i C i) is called a diagonal block
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 249
A Ferligoj Cluster Analysis 1
Outline1 Introduction 1
2 Cluster clustering blocks 2
4 Equivalences 4
9 Establishing blockmodels 9
11 Indirect approach 11
19 Direct approach 19
24 Generalized blockmodeling 24
26 Pre-specified blockmodeling 26
31 Symmetric-acyclic blockmodel 31
34 Blockmodeling of vauled networks 34
35 Blockmodeling of two-mode network 35
39 Blockmodeling of three-way networks 39
42 Application Clustering of Slovenian sociologists 42
48 Open problems in blockmodeling 48
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 349
A Ferligoj Cluster Analysis 1
Introduction
The goal of blockmodeling is to reduce a
large potentially incoherent network to a
smaller comprehensible structure that can be
interpreted more readily Blockmodeling
as an empirical procedure is based on the
idea that units in a network can be grouped
according to the extent to which they are
equivalent according to some meaningful
definition of equivalence (structural (Lorrain
and White 1971) regular (White and Re-
itz 1983) generalized (Doreian Batagelj
Ferligoj 2005)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 449
A Ferligoj Cluster Analysis 2
Cluster clustering blocksOne of the main procedural goals of blockmodeling is to identify in a given network
N = (U R) R sube U times U clusters (classes) of units that share structural charac-
teristics defined in terms of R The units within a cluster have the same or similar
connection patterns to other units They form a clustering C = C 1 C
2 C
k
which is a partition of the set U Each partition determines an equivalence relation
(and vice versa) Let us denote by sim the relation determined by partition C
A clustering C partitions also the relation R into blocks
R(C i C j) = R cap C i times C j
Each such block consists of units belonging to clusters C i and C j and all arcs leading
from cluster C i to cluster C j If i = j a block R(C i C i) is called a diagonal block
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 349
A Ferligoj Cluster Analysis 1
Introduction
The goal of blockmodeling is to reduce a
large potentially incoherent network to a
smaller comprehensible structure that can be
interpreted more readily Blockmodeling
as an empirical procedure is based on the
idea that units in a network can be grouped
according to the extent to which they are
equivalent according to some meaningful
definition of equivalence (structural (Lorrain
and White 1971) regular (White and Re-
itz 1983) generalized (Doreian Batagelj
Ferligoj 2005)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 449
A Ferligoj Cluster Analysis 2
Cluster clustering blocksOne of the main procedural goals of blockmodeling is to identify in a given network
N = (U R) R sube U times U clusters (classes) of units that share structural charac-
teristics defined in terms of R The units within a cluster have the same or similar
connection patterns to other units They form a clustering C = C 1 C
2 C
k
which is a partition of the set U Each partition determines an equivalence relation
(and vice versa) Let us denote by sim the relation determined by partition C
A clustering C partitions also the relation R into blocks
R(C i C j) = R cap C i times C j
Each such block consists of units belonging to clusters C i and C j and all arcs leading
from cluster C i to cluster C j If i = j a block R(C i C i) is called a diagonal block
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 449
A Ferligoj Cluster Analysis 2
Cluster clustering blocksOne of the main procedural goals of blockmodeling is to identify in a given network
N = (U R) R sube U times U clusters (classes) of units that share structural charac-
teristics defined in terms of R The units within a cluster have the same or similar
connection patterns to other units They form a clustering C = C 1 C
2 C
k
which is a partition of the set U Each partition determines an equivalence relation
(and vice versa) Let us denote by sim the relation determined by partition C
A clustering C partitions also the relation R into blocks
R(C i C j) = R cap C i times C j
Each such block consists of units belonging to clusters C i and C j and all arcs leading
from cluster C i to cluster C j If i = j a block R(C i C i) is called a diagonal block
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 649
A Ferligoj Cluster Analysis 4
Equivalences
Regardless of the definition of equivalence used there are two basic approaches to the
equivalence of units in a given network (compare Faust 1988)
bull the equivalent units have the same connection pattern to the same neighbors
bull the equivalent units have the same or similar connection pattern to (possibly)
different neighbors
The first type of equivalence is formalized by the notion of structural equivalence and
the second by the notion of regular equivalence with the latter a generalization of the
former
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 749
A Ferligoj Cluster Analysis 5
Structural equivalence
Units are equivalent if they are connected to the rest of the network in identical ways
(Lorrain and White 1971) Such units are said to be structurally equivalent
In other words X and Y are structurally equivalent iff
s1 XRY hArr YRX s3 forallZ isin U XY (XRZ hArr YRZ)
s2 XRX hArr YRY s4 forallZ isin U XY (ZRX hArr ZRY)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 849
A Ferligoj Cluster Analysis 6
Structural equivalence
The blocks for structural equivalence are null or complete with variations on diagonal
in diagonal blocks
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 949
A Ferligoj Cluster Analysis 7
Regular equivalence
Integral to all attempts to generalize structural equivalence is the idea that units are
equivalent if they link in equivalent ways to other units that are also equivalent
White and Reitz (1983) The equivalence relation asymp on U is a regular equivalence on
network N = (U R) if and only if for all XYZ isin U X asymp Y implies both
R1 XRZ rArr existW isin U (YRW and W asymp Z)
R2 ZRX rArr existW isinU
(WRY and W asymp Z)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1049
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1149
A Ferligoj Cluster Analysis 9
Establishing blockmodels
The problem of establishing a partition of units in a network in terms of a selected
type of equivalence is a special case of clustering problem that can be formulated as
an optimization problem (Φ P ) as follows
Determine the clustering C isin Φ for which
P (C) = minCisinΦ
P (C)
where Φ is the set of feasible clusterings and P is a criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1249
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1349
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1549
A Ferligoj Cluster Analysis 13
Example Support network among informatics students
The analyzed network consists of social support exchange relation among fifteen
students of the Social Science Informatics fourth year class (20022003) at the Faculty
of Social Sciences University of Ljubljana Interviews were conducted in October
2002Support relation among students was identified by the following question
Introduction You have done several exams since you are in the second class
now Students usually borrow studying material from their colleagues
Enumerate (list) the names of your colleagues that you have most oftenborrowed studying material from (The number of listed persons is not
limited)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1649
A Ferligoj Cluster Analysis 14
Class network - graph
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
classnet
Vertices represent students
in the class circles ndash girls
squares ndash boys Recipro-
cated arcs are represented by
edges
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1749
A Ferligoj Cluster Analysis 15
Class network ndash matrix
Pajek - shadow [000100]
b02
b03
g07
g09
g10g12
g22
g24
g28
g42
b51
g63b85
b89
b96
b 0 2
b 0 3
g 0 7
g 0 9
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
b 5 1
g 6 3
b 8 5
b 8 9
b 9 6
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1849
A Ferligoj Cluster Analysis 16
Indirect approach
b51
b89
b02
b96
b03
b85
g10
g24
g09
g63
g12
g07
g28
g22
g42
Using Corrected Euclidean-likedissimilarity and Ward clustering
method we obtain the following
dendrogram
From it we can determine the num-
ber of clusters lsquoNaturalrsquo cluster-
ings correspond to clear lsquojumpsrsquo in
the dendrogram
If we select 3 clusters we get the
partition C
C = b51 b89 b02 b96 b03 b85 g10 g24
g09 g63 g12 g07 g28 g22 g42
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 1949
A Ferligoj Cluster Analysis 17
Partition into three clusters (Indirect approach)
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
On the picture ver-
tices in the same
cluster are of the
same color
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2049
A Ferligoj Cluster Analysis 18
MatrixPajek - shadow [000100]
b02
b03
g10
g24 C1
b51
b85
b89
b96
g07
g22 C2
g28
g42
g09
g12 C3
g63
b 0 2
b 0 3
g 1 0
g 2 4
C 1
b 5 1
b 8 5
b 8 9
b 9 6
g 0 7
g 2 2
C 2
g 2 8
g 4 2
g 0 9
g 1 2
C 3
g 6 3
The partition can be used
also to reorder rows andcolumns of the matrix repre-
senting the network Clus-
ters are divided using blue
vertical and horizontal lines
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2149
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2249
A Ferligoj Cluster Analysis 20
Criterion function
One of the possible ways of constructing a criterion function that directly reflectsthe considered equivalence is to measure the fit of a clustering to an ideal one with
perfect relations within each cluster and between clusters according to the considered
equivalence
Given a clustering C = C 1 C 2 C k let B (C u C v) denote the set of all ideal
blocks corresponding to block R(C u C v) Then the global error of clustering C can
be expressed as
P (C) =
C uC visinC
minBisinB(C uC v)
d(R(C u C v) B)
where the term d(R(C u C v) B) measures the difference (error) between the block
R(C u C v) and the ideal block B d is constructed on the basis of characterizations
of types of blocks The function d has to be compatible with the selected type of
equivalence
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2349
A Ferligoj Cluster Analysis 21
Empirical blocks
a b c d e f g
a 0 1 1 0 1 0 0b 1 0 1 0 0 0 0
c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0f 1 1 1 0 1 0 1g 0 1 1 0 0 0 0
Ideal blocks
a b c d e f ga 0 1 1 0 0 0 0
b 1 0 1 0 0 0 0c 1 1 0 0 0 0 0
d 1 1 1 0 0 0 0e 1 1 1 0 0 0 0
f 1 1 1 0 0 0 0g 1 1 1 0 0 0 0
Number of inconsistencies
for each block
A B
A 0 1B 1 2
The value of the criterion function is the sum of all incon-sistencies P = 4
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2449
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2549
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2649
A Ferligoj Cluster Analysis 24
Generalized blockmodeling
1 1 1 1 1 1 0 0
1 1 1 1 0 1 0 1
1 1 1 1 0 0 1 0
1 1 1 1 1 0 0 0
0 0 0 0 0 1 1 1
0 0 0 0 1 0 1 1
0 0 0 0 1 1 0 1
0 0 0 0 1 1 1 0
C 1 C 2
C 1 complete regular
C 2 null complete
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2749
A Ferligoj Cluster Analysis 25
Generalized equivalence block types
Y 1 1 1 1 1
X 1 1 1 1 11 1 1 1 11 1 1 1 1
complete
Y 0 1 0 0 0
X 1 1 1 1 10 0 0 0 00 0 0 1 0
row-dominant
Y 0 0 1 0 0
X 0 0 1 1 01 1 1 0 00 0 1 0 1
col-dominant
Y 0 1 0 0 0
X 1 0 1 1 00 0 1 0 11 1 0 0 0
regular
Y 0 1 0 0 0
X 0 1 1 0 01 0 1 0 00 1 0 0 1
row-regular
Y 0 1 0 1 0
X 1 0 1 0 01 1 0 1 10 0 0 0 0
col-regular
Y 0 0 0 0 0
X 0 0 0 0 00 0 0 0 00 0 0 0 0
null
Y 0 0 0 1 0
X 0 0 1 0 01 0 0 0 00 0 0 1 0
row-functional
Y
1 0 0 00 1 0 0X 0 0 1 0
0 0 0 00 0 0 1
col-functional
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2849
A Ferligoj Cluster Analysis 26
Pre-specified blockmodeling
In the previous slides the inductive approaches for establishing blockmodels for a
set of social relations defined over a set of units were discussed Some form of
equivalence is specified and clusterings are sought that are consistent with a specifiedequivalence
Another view of blockmodeling is deductive in the sense of starting with a blockmodel
that is specified in terms of substance prior to an analysis
In this case given a network set of types of ideal blocks and a reduced model asolution (a clustering) can be determined which minimizes the criterion function
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 2949
A Ferligoj Cluster Analysis 27
Types of pre-specified blockmodels
The pre-specified blockmodeling starts with a blockmodel specified in terms of
substance prior to an analysis Given a network a set of ideal blocks is selected a
family of reduced models is formulated and partitions are established by minimizing
the criterion function
The basic types of models are
0 0
0 0
0 0
0
0 0
0 0
0 0
core - hierarchy clustering
periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3049
A Ferligoj Cluster Analysis 28
Pre-specified blockmodeling example
We expect that core-periphery model exists in the network some students having
good studying material some not
Prespecified blockmodel (comcomplete regregular -null block)
1 21 [com reg] -
2 [com reg] -
Using local optimization we get the partition
C = b02 b03 b51 b85 b89 b96 g09
g07 g10 g12 g22 g24 g28 g42 g63
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3149
A Ferligoj Cluster Analysis 29
2 Clusters Solution
b02
b03
g07
g09
g10
g12
g22 g24
g28
g42
b51
g63
b85
b89
b96
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3249
A Ferligoj Cluster Analysis 30
Model
Pajek - shadow [000100]
g07
g10
g12
g22
g24
g28
g42
g63
b85
b02
b03
g09
b51
b89
b96
g 0 7
g 1 0
g 1 2
g 2 2
g 2 4
g 2 8
g 4 2
g 6 3
b 8 5
b 0 2
b 0 3
g 0 9
b 5 1
b 8 9
b 9 6
Image and Error Matrices
1 21 reg -
2 reg -
1 21 0 3
2 0 2
Total error = 5
center-periphery
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3349
A Ferligoj Cluster Analysis 31
Symmetric-acyclic blockmodel
Let us introduce a new type of permitted diagonal block and a new type of blockmodel
respectively labeled the lsquosymmetricrsquo block and the lsquoacyclic blockmodelrsquo The idea
behind these two objects comes from a consideration of the idea of a ranked clusters
model from Davis and Leinhardt (1972) We examine the form of these models
and approach them with the perspective of pre-specified blockmodeling (DoreianBatagelj Ferligoj 2000 2005)
We say that a clustering C = C 1 C k over the relation R is a symmetric-acyclic
clustering iff
bull all subgraphs induced by clusters from C (diagonal blocks) contain onlybidirectional arcs (edges) and
bull the remaining graph (without these subgraphs) is acyclic (hierarchical)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3449
A Ferligoj Cluster Analysis 32
An example of symmetric-acyclic blockmodel
The pre-specified blockmodel is acyclic with symmetric diagonal blocks An example
of a clustering into 3 clusters symmetric-acyclic type of models can be specified in the
following way
sym nul nul
nulone sym nul
nulone nulone sym
The criterion function for a symmetric-acyclic blockmodel is simply a count of the
number of (1) asymmetric ties in diagonal blocks and (2) non-zero ties above the
diagonal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3549
A Ferligoj Cluster Analysis 33
Table 1 An Ideal symmetric-acyclic network
a 1 1 1 b 1 1 1 1 c 1 1 1 1 d 1 1 1 1 e 1 1 1 f 1 1 1 1 1 1 g 1 1 1 1 1 h 1 1 1 1 1 1
i 1 1 1 1 1 1 1 j 1 1 1 1 1 1 k 1 1 1 1 1 1 l 1 1 1 1 1 m 1 1 1 1 1 1 1 1 n 1 1 1 1 1 1 1 1 o 1 1 1 1 1 1 1 1 1 p 1 1 1 1 1 1 1 1
q 1 1 1 1 1 1 1 1 r 1 1 1 1 1 1 1s 1 1 1 1 1 1 1 1 1 1t 1 1 1 1 1 1 1 1u 1 1 1 1 1 1 1 1 1 1v 1 1 1 1 1 1 1 1 1 1 1
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3649
A Ferligoj Cluster Analysis 34
Blockmodeling of vauled networks
Another interesting problem is the development of blockmodeling of valued networks (BatageljFerligoj 2000 Ziberna 2007)
Ziberna (2007) proposed several approaches to
generalized blockmodeling of valued networks
where values of the ties are assumed to be mea-
sured on at least interval scaleThe first approach is a straightforward gen-
eralization of the generalized blockmodeling
of binary networks (Doreian P Batagelj V
Ferligoj A 2005) to valued blockmodeling
The second approach is homogeneity blockmod-eling The basic idea of homogeneity blockmod-
eling is that the inconsistency of an empirical
block with its ideal block can be measured by
within block variability of appropriate values
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3749
A Ferligoj Cluster Analysis 35
Blockmodeling of two-mode network
It is also possible to formulate a generalized blockmodeling problem where thenetwork is defined by two sets of units and ties between them
Therefore two partitions mdash row-partition and column-partition have to be determined
Two-mode generalized blockmodeling can be done
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3849
A Ferligoj Cluster Analysis 36
Supreme Court voting for twenty-six important decisions
Issue Label Br Gi So St OC Ke Re Sc Th
Presidential Election PE - - - - + + + + +Criminal Law Cases
Illegal Search 1 CL1 + + + + + + - - -Illegal Search 2 CL2 + + + + + + - - -Illegal Search 3 CL3 + + + - - - - + +Seat Belts CL4 - - + - - + + + +Stay of Execution CL5 + + + + + + - - -
Federal Authority Cases
Federalism FA1 - - - - + + + + +Clean Air Action FA2 + + + + + + + + +Clean Water FA3 - - - - + + + + +
Cannabis for Health FA4 0 + + + + + + + +United Foods FA5 - - + + - + + + +NY Times Copyrights FA6 - + + - + + + + +
Civil Rights Cases
Voting Rights CR1 + + + + + - - - -Title VI Disabilities CR2 - - - - + + + + +PGA v Handicapped Player CR3 + + + + + + + - -
Immigration Law Cases
Immigration Jurisdiction Im1 + + + + - + - - -Deporting Criminal Aliens Im2 + + + + + - - - -Detaining Criminal Aliens Im3 + + + + - + - - -Citizenship Im4 - - - + - + + + +
Speech and Press CasesLegal Aid for Poor SP1 + + + + - + - - -Privacy SP2 + + + + + + - - -Free Speech SP3 + - - - + + + + +Campaign Finance SP4 + + + + + - - - -Tobacco Ads SP5 - - - - + + + + +
Labor and Property Rights Cases
Labor Rights LPR1 - - - - + + + + +Property Rights LPR2 - - - - + + + + +
The Supreme Court Justices and
their lsquovotesrsquo on a set of 26 ldquoimpor-
tant decisionsrdquo made during the
2000-2001 term Doreian and Fu-
jimoto (2002)
The Justices (in the order in
which they joined the Supreme
Court) are Rehnquist (1972)
Stevens (1975) OrsquoConner (1981)
Scalia (1982) Kennedy (1988)
Souter (1990) Ginsburg (1993)
and Breyer (1994)
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 3949
A Ferligoj Cluster Analysis 37
Supreme Court voting a (47) partition
C am p ai gnF i
V o t i n gR i gh
D e p or t A l i e
I l l e g al S 1
I l l e g al S 2
S t a y E x e c u t
P GA v P l a y er
P r i v a c y
I mmi gr a t i o
L e g al A i d P o
C r i mi n al A l
I l l e g al S 3
C l e anA i r A c
C ann a b i s
N Y T i m e s
S e a t B el t
U ni t e d F o o d
C i t i z en s h i
P r e s E l e c t i
F e d er al i s m
F r e e S p e e c h
C l e anW a t er
T i t l eI V
T o b a c c o
L a b or I s s u e
P r o p er t y
Rehnquest
Thomas
Scalia
Kennedy
OConner
Breyer
Ginsburg
Souter
Stevens
upper ndash conservative lower ndash liberal
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4049
A Ferligoj Cluster Analysis 38
Blockmodeling of three-way networks
The indirect approach to structural equivalence blockmodeling in 3-mode networks
Indirect means ndash embedding the notion of equivalence in a dissimilarity and deter-
mining it using clustering A 3-mode network N over the basic sets X Y and Z is
determined by a ternary relation R sube X timesY timesZ The notion of structural equivalencedepends on which of the sets X Y and Z are (considered) the same There are three
basic cases
bull all three sets are different
bull two sets are the same
bull all three sets are the same
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4149
A Ferligoj Cluster Analysis 39
Example Artificial dataset
Randomly generated ideal structure rndTest(c(564)c(353535))
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4249
A Ferligoj Cluster Analysis 40
Example Solutions
1 9 b f i k o 3 4 d e j l m u v w q r s 2 5 7 a c g t x z 6 8 h n p y
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 1) method = ward)
H e
i g h t
1 3 4 9 f m y 2 7 a d r 8 g j n w x 5 b p t z c e i l o s v 6 h k q u
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 2) method = ward)
H e
i g h t
1 4 5 6 7 8 a b c q r u 3 9 g j k l o s v x 2 d e f h m p y z i n t w
0
5 0 0
1 0 0 0
1 5 0 0
2 0 0 0
Dendrogram of agnes(x = dist3m(t 0 3) method = ward)
H e
i g h t
Budapest September 16 2009
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4349
A F li j Cl A l i 42
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4449
A Ferligoj Cluster Analysis 42
Indirect blockmodeling approach
First the indirect approach for structural equivalence for coauthorship network wasperformed The Ward dendrogram where the corrected euclidean-like dissimilarity
was used is presented in the dendrogram
Budapest September 16 2009
A F li j Cl t A l i 43
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4549
A Ferligoj Cluster Analysis 43
Pre-specified blockmodeling
In the case of the coauthorship network we can assume a core-periphery structure one
or several clusters of strongly connected scientists that strongly collaborate among
themselves The last cluster consists of scientists that do not collaborate among
themselves and also do not with the scientists of the other clusters They publishby themselves or with the researchers from the other scientific disciplines or with
researchers outside of Slovenia
We performed the core-periphery pre-specified blockmodels (structural equivalence)
into two to ten clusters The highest drop of the criterion function was obtained for
the blockmodels into three five and eight clusters As eight clusters have been shown
also by the dendrogram this clustering is further analyzed
Budapest September 16 2009
A F li j Cl t A l i 44
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4649
A Ferligoj Cluster Analysis 44
Core-periphery structure of the coauthorship network
Budapest September 16 2009
A Ferligoj Cluster Analysis 45
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4749
A Ferligoj Cluster Analysis 45
If we permute rows and columns in the relational matrix in such a way that first the
units of the first cluster are given than the units of the second cluster and so on and
if we cut the relational matrix by clusters we obtain the table in the next slide Here
a black square is drawn if the two researchers have at least one joint publication or a
white square if they have not publish any joint publication
In the table the obtained blockmodel into eight clusters is presented (seven core
clusters and one peripherical one) The seven core clusters are rather small ones and
the peripherical large one
Only cluster 4 is connected also to the first and second clusters all other core clusters
show cohesivness inside the cluster and nonconnectivity outside the cluster
The first two diagonal blocks are complete without any inconsistency compared tothe ideal complete blocks All white squares in complete blocks and black squares in
zero blocks are inconsistencies The number of all inconsistencies is the value of the
criterion function For the blockmodel the value of the criterion function is 310
Budapest September 16 2009
A Ferligoj Cluster Analysis 46
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4849
A Ferligoj Cluster Analysis 46
Main results
bull There are several small core groups of the Slovenian sociologists that publishamong themselves and much less or not at all with the members of the other
groups
bull The core groups overlap with the organizational structure sociologists of each of
these core groups are members of a research center
bull The periphery group composed by the sociologists that are mostly not coauthors
with the other Slovenian sociologists is very large (more than 56 ) Most of
them publish only as a single author Probably some of them publish with the
other Slovenian nonsociologists or with the researchers outside of Slovenia
bull The coauthorship network is quite a sparse one which tells us that the preferred
publication culture by Slovenian sociologists is to publish as a single author
bull More detailed analysis of the periphery group should be further studied
Budapest September 16 2009
A Ferligoj Cluster Analysis 47
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)
8112019 Doreian Batagelj Ferligoj - Generalized Blockmodeling
httpslidepdfcomreaderfulldoreian-batagelj-ferligoj-generalized-blockmodeling 4949
A Ferligoj Cluster Analysis 47
Open problems in blockmodeling
The current local optimization based programs for generalized blockmodeling can
deal only with networks with at most some hundreds of units What to do with larger
networks is an open question For some specialized problems also procedures for
(very) large networks can be developed (Doreian Batagelj Ferligoj 2000 Zaversnik
Batagelj 2004)