1 Reliability and Validity Research Methods in AD/PR COMM 420.8 Fall 2007 Nan Yu.

31
1 Reliability and Validity Research Methods in AD/PR COMM 420.8 Fall 2007 Nan Yu

Transcript of 1 Reliability and Validity Research Methods in AD/PR COMM 420.8 Fall 2007 Nan Yu.

1

Reliability and Validity

Research Methods in AD/PRCOMM 420.8

Fall 2007Nan Yu

2

Exam 1

Multiple choices (70%)Short answer (30%)Time: 9/27 (Thursday), 3:35-5:30pPlace: 143 StuckemanNo made-up exams will be given if you

miss the exam without a prior notice or a verifiable excuse.

3

Overview of Last Class

Levels of measurementNominalOrdinalIntervalRatio

Types of measurementOpen-ended questionLikert-type scaleThurstone scaleSemantic Differential scale

4

How to differentiate interval/ratio/ordinal variable? How many hours do you watch TV everyday? ________hours

How many hours do you watch TV everyday? 1 2 3 4 5 6 7 8 9hour

hours

How many hours do you watch TV every day?• 0-2• 3-5• 5-7• 7-9• 9-11• more than 11 hours

5

Nominal variable

Categories are mutually exclusive

Categories are exhaustive: all possible responses are provided (One individual case should fit in at least one category.)Ethnicity

• Caucasian• African American• Native American• Asian• Hispanic• Other

6

Likert-type scale

Likert-type scaleA neutral point is always provided

Listening to heavy metal music makes one prone to violent acts. __Strongly agree __Agree __Neutral __Disagree __Strongly disagree

How would you rate the quality of Daily Collegian? __poor__ unsatisfactory __neither unsatisfactory nor satisfactory __ satisfactory __ excellent

Listening to heavy metal music makes one prone to violent acts.Strongly disagree Neutral Strongly agree

-3 -2 -1 0 1 2 3

7

Semantic Differential Scales (bipolar)

If the scale looks like this: Not at all Very much

Likable 1 2 3 4 5 6 7 8 9Good 1 2 3 4 5 6 7 8 9Pleasant 1 2 3 4 5 6 7 8 9

Then what type of measurement is it now?

8

Corrections

http://www.personal.psu.edu/mbo1/forms/survey1/survey1.html

A few corrections0-25% 26%-50% 51%-75% 76-100%

• ordinal, closed question

Gender 1=male, 2=female• nominal, closed question

9

Reliability

The extend to which measurement are consistent, stable, dependable, predicableSuggests that the same thing is repeated or recurs under

identical or similar conditionsEvery time you measure, you get similar or same data

Why reliability is important?Guarantee the quality of your dataReplicability

E.g., degree in which replicating a study using the same procedures, the same instruments, etc. will lead to the same results

10

Factors that reduce reliability

Instrumental errorDouble-barreled question

Do you like to watch and play basketball?

Application errorInstrument is used improperly

Random errorUnpredictable error

11

Types of Reliability

Test-retest reliability Degree of matching between measurement results

when the measurements are repeated E.g., They are taken more than once, for the same

object of measurement

12

Types of Reliability

Measurement item reliability Internal consistency of several itemsThe degree in which a bunch of items stand together.

E.g., happiness measured by answering questions such as • “how thrilled you are?” 1 2 3 4 5

• “how happy you are?” 1 2 3 4 5

• “how cheerful you are?” 1 2 3 4 5

Answers to these three questions should be similar; it would mean that the happiness scale is reliable / it has internal consistency

Cronbach alpha: 0.7-0.9 (interval variable)

13

Types of Reliability

Inter-coder reliabilityWhen a researcher has to code or interpret open-

ended answers of the respondents, or news stories material, etc., his or her interpretation might be subjective and therefore, not completely reliable.

One way of dealing with this problem is to ask other individuals to “code” the same material.

Degree in which they agree upon the results of coding is inter-coder reliability

90% of agreement, Cohen’s kappa, Scott’s pi.

14

Inter-coder reliability example

Imagine that three coders are asked to code the amount of violence on a certain televised program; they are given a coding sheet, explaining to them what they should consider as being violent. However, they do not always agree that certain acts or behaviors pertain to one of the violence descriptions. If they agree 87 percent of the time, the inter-coder reliability is of 87% (0.87).

15

How to improve reliability

Clearly conceptualize all constructs; reliability increases when a single construct or sub-dimension of a construct is measured

Increase the level of measurement; more precise levels of measurement are more likely to be reliable than less precise measures because the latter pick up less detailed information

Age• Young• Middle-aged• Old

Age0-10 11-20 21-30

31-40 41-50 51-60 61+

Age• What is your age?__________

16

How to improve reliability

Use multiple indicators of a variable; multiple indicator measures tend to be more stable than measures with one item

Sadness• gloom• sorrow• grief• unhappiness

Use pretests, pilot studies, and replicationtrained observers/coders

17

Validity

Degree to which a measure “measures” what is supposed to measure (e.g., degree of matching between the concept and the measurement).

18

Content and Face Validity

Whether a measure captures the meaning of the variable being measured.

19

Face and Content Validity

Face validity In face validity, you look at the operationalization and see whether "on

its face" it seems like a good translation of the construct. Example of lacking face validity

Use a ruler to measure weight

Use shoe size to measure intelligence

Content validity Very similar to face validity

Needs careful operationalization of the concept.

Researchers are the judges of measurement validity (face and content).

20

Criterion Validity

Criterion validityUses some standard or criterion to indicate a

construct accurately.“concurrent validity” and “predictive validity”

21

Types of Criterion Validity

Concurrent validity:

a measure (indicator) must be associated with a preexistent one that is judged to be valid.

• E.g. GRE

Predictive validity: indicator predicts future events that are logically

related to a construct is called predictive validity.• E.g. SAT and college academic performance

22

Construct validity

Construct validity refers to the degree to which inferences can

legitimately be made from the operationalizations in your study to

the theoretical constructs on which those operationalizations were

based.

Convergent validity

Discriminant validity (Divergent validity)

23

Construct validity

Convergent validity (multiple-item measures)Applies when multiple indicators converge or are associated

with one another.

you should be able to show a correspondence or convergence

between similar constructs

Convergent validity means that multiple measures of the same

construct hang together or operate in similar ways.

24

Convergent validity (multi-item validity)

25

Discriminant validity

Discriminant validitymeasures of constructs that theoretically should not be

related to each other are, in fact, observed to not be related

to each other

you should be able to discriminate between dissimilar

constructs

you are measuring what you want to measure, not

something else.

26

Discriminant validity

27

Internal Validity

Internal validityDegree to which one can prove causation

Practically, degree to which you can eliminate third variables or confounds.

28

External Validity

External validityDegree to which one can generalize the conclusions of the study from the sample used in the study to the overall population.

generalize

29

Relationship betweenreliability and validity (p. 143-144)

“Reliability must be present or validity is impossible.”Reliability is a necessary condition for validity, but not

sufficient.Even the measure turns to be reliability, it may not

measure what you want to measure --lack of validity

30

Relationship betweenreliability and validity

31

In-Class Demo 1 and 2

Download the file “in-class demo 1” and “1-class demo 2” in “week 5” folder on ANGEL

Complete both of themSubmit your answers to the corresponding

drop boxes in “week 5” folder on ANGELAnswers of are also in the “week 5” folder.