Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole...

36
Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student, NUS) Shengdong (Shen) Zhao (Faculty member, NUS) Michael J. McGuffin 1 (Faculty member, École de technologie supérieure)

Transcript of Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole...

Page 1: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Tracing Tuples Across Dimensions

A Comparison of Scatterplots and Parallel Coordinate Plots

Xiaole Kuang (Master student, NUS)

Haimo Zhang (PhD student, NUS)

Shengdong (Shen) Zhao (Faculty member, NUS)

Michael J. McGuffin

1

(Faculty member, École de technologie supérieure)

Page 2: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

2

The Last Talk of The Last Session of

The Last Day!

Welcome to

Page 3: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

3

of

Vienna

Singapore

9697 km

Page 4: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

4

Vignette (CHI ‘12)SandCanvas (CHI ‘11)MOGCLASS (CHI ‘11)Magic Cards (CHI ‘09)

earPod (CHI ‘07)Zone & Polygon Menu (CHI ‘06)

Elastic Hierarchy (InfoVis ‘05)

Simple Marking Menu (UIST ‘04)

Systems, Tools, Interaction Techniques

Page 5: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Visualization Techniques for Multi-Variate Data

Scatter Plot (SCP)

Parallel Coordinate Plot

(PCP)

Scatter Plot Matrix

(SPLOM)

5

Page 6: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Why PCP vs. SCP?Both techniques are popular!Yet, we know very little about their comparative advantages.

6

Viau et al., TVGC10

Yuan et al., TVGC09

Claessen & van Wijk, TVGC11

We need more systematic evaluations between PCP & SCP!

Page 7: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Basics of Evaluation Research question• What’s the comparative advantages

between PCP & SCP for certain tasks?

Task Independent variablesDependent variables

7

Page 8: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Basics of Evaluation Research question• What’s the comparative advantages

between PCP & SCP for certain tasks?

Task Independent variablesDependent variables

8

Page 9: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Basic Analytical Tasks

9

serves as a subtask for many other tasks

Amar et al.: Low-level components of analytic activity in information visualization. InfoVis05, 111–117.

(Holten & van Wijk, EuroVis10)

(Li et al., InfoVis10)

PCP is inferior than SCP

Page 10: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Value Retrieval TaskDefinition: • Given the numerical value of one attribute

of a data tuple, find the numerical value of another attribute of the same data tuple.

10

Multi-Variate Data Tuple (X1, X2, X3, …. , Xn)

a ?

Page 11: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Basics of Evaluation Research question• What’s the comparative advantages

between PCP & SCP for certain tasks?

Task Independent variablesDependent variables

11

Page 12: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Independent Variables

12

Technique

Parallel Coordinate Plot (PCP)

Scatter Plot (SCP)

X2

X1

X3

X2

X4

X3

X1

X2

X3

X2

X4

X2

X 2X 1

X 2X 3

X 4X 3

SCP-rotated (Qu et al., TVCG07)

SCP-common (SPLOM)

SCP-staircase (Viau et al., TVCG10)

Page 13: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Independent Variable – 4 Technique

13

PCP

SCP-common(i.e., SPLOM)

SCP-rotated(i.e., Qu et al., TVCG07)

SCP-staircase(i.e., Viau et al., TVCG10)

Page 14: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Additional Independent Variables

14

X2

X1

X3

X2

X4

X3

Number of Dimensions

X2

X1

X3

X2

X4

X3

X5

X4

Data Density

X2

X1

X3

X2

X4

X3

Page 15: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Independent Variables• Technique• Dimension• Density

15

Page 16: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Dependent Variables• Completion time• Error distance

16

Page 17: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Experiment Demo

17

Page 18: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Experiment 1 Design 12 participants

× 4 visualization techniques (PCP, SCP-common, SCP-rotate, SCP-standard)

× 3 levels of data dimension (2D, 4D, 6D)

× 3 levels of data density (10 tuples, 20 tuples, 30 tuples)

× 3 repetitions of trials

= 1296 trials in total.

18

Page 19: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Secon

ds

SCP-rotate

SCP-common SCP-staircase

PCP

Overall Results

19

Best Good

Poor

Completion Time Error Distance

Err

or

Dis

tan

ce

SCP-rotate

SCP-common SCP-staircase

PCP

Poor

Good

Poorer

Page 20: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

1st Take-away Lesson

20

PCP

SCP-common(i.e., SPLOM)

SCP-rotated(i.e., Qu et al., TVCG07)

SCP-staircase(i.e., Viau et al., TVCG10)

Page 21: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

PCP vs. SCP-common

21

Page 22: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

PCP vs. SCP-common

22

Density

Performance Difference

Page 23: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

PCP vs. SCP-common

23

Density

Performance Switch Order

Page 24: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Important Observation

There seems to be a

Density & Number of Dimension Trade-off

between PCP & SCP-common!

24

Page 25: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Experiment 2× 18 participants

× 2 techniques (PCP, SCP-common)

× 3 dimensions (4D, 6D, 8D) [2D, 4D, 6D in Exp. 1]

× 3 densities (20 tuples, 30 tuples, 40 tuples) [10, 20, 30 in Exp. 1]× 5 trials for each combination

= 1620 trials in total.25

Page 26: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Results – Completion Time

26

Overall result for Exp. 2

SCP-common (15.41s) PCP (18.23s)

Result in Exp. 1

SCP-common (12.02s) PCP (8.99s)

faster

faster

Trade-off between number of dimensions & data density

Dimension Density

Page 27: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Results – Error Distance

27

Trade-off between number of dimensions & data density

Dimension Density

Page 28: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Take-away LessonsThe value retrieval performance of PCP increases depending on dimensionality.The performance of SCP-common seems independent of dimensionality.

Increasing density affects the performance of PCP more than it affects SCP-common.

28

Dimension

Density

Page 29: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Let’s Recap the Take Away-Messages and Ask

Why1) Both SCP-rotate and SCP-staircase are inferior for value retrieval task

29

Page 30: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Let’s Recap the Take Away Messages

2) Performance trade-off between PCP & SCP-common for both dimensionalities and data density.• PCP increases depending

on dimensionality. • SCP-common

performance seems to be independent.

30

Page 31: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Let’s Recap the Take Away Messages

31

10 tuples

40 tuples

2) Performance trade-off between PCP & SCP-common for both dimensionalities and data density.• PCP increases depending

on dimensionality. • SCP-common

performance seems to be independent.

• Increasing density affects the performance of PCP more than it affects SCP-common.

Page 32: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Conclusion and Future Work

Our study helps to understand the comparative advantages between PCP & SCP

However, this is only a starting point,

32

Page 33: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

The Grand VisionIdeally, this problem can be solved by …

33

InfoVisevaluation

package

Results/Recommendations

Page 34: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

AcknowledgmentThis research is supported by:

The National University of Singapore Academic Research Fund R-252-000-375-133

and by:

The Singapore National Research Foundation under its International Research Centre @ Singapore Funding Initiative and administered by the IDM Programme Office.

Page 35: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

Q & A

35

Elastic Hierarchy (InfoVis ‘05)

Tracing Tuples Across Dimensions (EuroVis ‘12)

Page 36: Tracing Tuples Across Dimensions A Comparison of Scatterplots and Parallel Coordinate Plots Xiaole Kuang (Master student, NUS) Haimo Zhang (PhD student,

End

36