© mcgraw-hill higher education. all rights reserved. chapter 4 validity
TRANSCRIPT
© McGraw-Hill Higher Education. All rights reserved.
Chapter 4Chapter 4
ValidityValidity
© McGraw-Hill Higher Education. All rights reserved.
Chapter 4 OutlineChapter 4 Outline
Understanding Validity Evidences of Validity Content Validity Evidence Evidence of Logical Validity Criterion-related Validity Evidence Factors that Affect Criterion-related Validity Construct Validity Evidence Criterion-referenced Test Validity
© McGraw-Hill Higher Education. All rights reserved.
ValidityValidity
Most important characteristic of a measurement.
© McGraw-Hill Higher Education. All rights reserved.
ValidityValidity
Extent to which the inferences made from specific measures are appropriate.
It is not the measures that are validated; it is the inferences that are validated.
© McGraw-Hill Higher Education. All rights reserved.
ValidationValidation
Process of obtaining evidence to demonstrate that inferences made from test scores are appropriate and meaningful.
© McGraw-Hill Higher Education. All rights reserved.
Evidence of validity:Evidence of validity:
judged based on the intended use of the data.should come from various sources.slowly evolves and allows us to place
confidence in the measurement instrument.
© McGraw-Hill Higher Education. All rights reserved.
ConstructConstruct
The object of interest.Examples: aerobic fitness, upper body
strength, physical activity, body composition, perceived competence.
All validation is “construct validation.”
© McGraw-Hill Higher Education. All rights reserved.
Evidences of ValidityEvidences of Validity
Content validity evidenceLogical validity evidence
Criterion-related validity evidence Construct validity evidence
© McGraw-Hill Higher Education. All rights reserved.
Content Validity EvidenceContent Validity Evidence
Logical approachThe term “content validity” is used with
written knowledge tests.Judgments of experts are used to determine
whether the items on the test represent all of the important content areas.
A property of the test rather than a property of the use of the test scores.
© McGraw-Hill Higher Education. All rights reserved.
Evidence of Logical Validity Evidence of Logical Validity
Concept of logical validity was developed for physical tests in the field of exercise science.
Extent to which a test measures the most important components of skill necessary to perform a motor task adequately.
© McGraw-Hill Higher Education. All rights reserved.
Criterion-related Criterion-related Validity EvidenceValidity Evidence
Demonstrated by correlating the test of the construct with a criterion measure of that construct.
Two types of criterion-related evidence designs:– concurrent evidence– predictive evidence
© McGraw-Hill Higher Education. All rights reserved.
Used when a test is proposed as a substitute for the criterion measure.
Crucial step: selection of criterion. Criteria typically come from:
Expert judgesTournament standingsKnown valid test
Concurrent Evidence of Validity
© McGraw-Hill Higher Education. All rights reserved.
Subjective ratings by one or more experts.Either point rating scale or group ranking.Point rating scale is preferable.Commonly used with skill tests.
Scores from the test are correlated with scores from the expert ratings.
Expert Ratings
© McGraw-Hill Higher Education. All rights reserved.
Assumes tournament standing is a good indicator of overall ability in the tested skill.
Usually used with an individual or dual activity (e.g., tennis, badminton).
Scores from the test are correlated with tournament standings.
Tournament Standings
© McGraw-Hill Higher Education. All rights reserved.
Predetermined criteria are scores from an accepted instrument.
Scores from the test are correlated with scores from the predetermined criteria.
Usually used to develop a quicker and easier test than criterion.e.g., skinfolds are used to evaluate body
composition. What’s the criterion?
Predetermined Criteria
© McGraw-Hill Higher Education. All rights reserved.
Criterion measure should be recognized as the “gold standard.”
Multiple correlations and regression equations are often used to establish evidence of concurrent validity.
Predetermined Criteria
© McGraw-Hill Higher Education. All rights reserved.
Used when want to predict the criterion measure some time in the future.
For Predictive Evidence the criterion is measured some time “in the future”.
For Concurrent Evidence the criterion is measured at about the same point in time.
Predictive Evidence of Validity
© McGraw-Hill Higher Education. All rights reserved.
Factors that Affect Criterion-Factors that Affect Criterion-related Validity Coefficientrelated Validity Coefficient
Criterion measure selected Characteristics of individuals tested Reliability
Maximum possible validity coefficient
rxy = (Rxx)(Ryy)
Objectivity Test length
© McGraw-Hill Higher Education. All rights reserved.
Size of Criterion-related Size of Criterion-related Validity CoefficientValidity Coefficient
Depends on:– criterion used– what others conducting similar studies have
found– accuracy needed
© McGraw-Hill Higher Education. All rights reserved.
Construct Validity EvidenceConstruct Validity Evidence
Based on scientific method.– Hypothesize about the construct– Develop theory to explain construct, its
relationships with other constructs, and/or the tests proposed to measure it
– Apply procedures to confirm or disconfirm the theory
Judge extent to which theoretical and statistical information supports the constructs.
© McGraw-Hill Higher Education. All rights reserved.
Some Methods for Confirmatory Some Methods for Confirmatory Evidence of ValidityEvidence of Validity
Factor Analysis – to evaluate internal evidence.
Multitrait, Multimethod Matrix – to obtain convergent and discriminant evidence.
Known Difference Method.
© McGraw-Hill Higher Education. All rights reserved.
Factor AnalysisFactor Analysis
Used to confirm the nature of the construct. Can determine whether items seem to
measure the same construct.
© McGraw-Hill Higher Education. All rights reserved.
Multitrait, Multimethod Matrix Multitrait, Multimethod Matrix (MTMM)(MTMM)
Convergent evidence – measures of the same construct correlate highly.
Discriminant evidence – measures of different constructs do not correlate as highly as measures of the same construct.
© McGraw-Hill Higher Education. All rights reserved.
Known Difference MethodKnown Difference Method
Compare scores of two different groups that are expected to differ.e.g., advanced vs. intermediate swimmerse.g., comparison of group before and after
training or instruction.
© McGraw-Hill Higher Education. All rights reserved.
New 3-StageNew 3-StageParadigm for ValidityParadigm for Validity
Definitional Evidence StageConfirmatory Evidence StageTheory Testing Stage
© McGraw-Hill Higher Education. All rights reserved.
Definitional Stage
Confirmatory Stage
Theory-
Testing
Stage
© McGraw-Hill Higher Education. All rights reserved.
Definitional Evidence StageDefinitional Evidence Stage
Investigation of evidence to describe the nature of the construct of interest.
Content or logical approach to collecting validity evidence is used at this stage.
This evidence differs from other approaches to validation in that properties of the test rather than interpretations of the test are examined.
© McGraw-Hill Higher Education. All rights reserved.
Confirmatory Evidence StageConfirmatory Evidence Stage
Collect evidence to confirm or disconfirm the description of the construct.
© McGraw-Hill Higher Education. All rights reserved.
Theory Testing StageTheory Testing Stage
Test theories of how the construct of interest fits into the broader context.considerations of how the construct is related to other
constructs.determinants of a person’s status on the construct of
interest.
© McGraw-Hill Higher Education. All rights reserved.
Criterion-referenced Test (CRT) Criterion-referenced Test (CRT) ValidityValidity
Criterion-referenced test -- test with a
predetermined standard used to classify people
as proficient or nonproficient.
Different techniques are used to validate CRT.
© McGraw-Hill Higher Education. All rights reserved.
Two Approaches to CRT ValidityTwo Approaches to CRT Validity
Logical -- Domain-referenced validity
Statistical -- Decision validity or decision
accuracy
© McGraw-Hill Higher Education. All rights reserved.
Domain-referenced ValidityDomain-referenced Validity
Domain -- behavior the test is designed to
measure.
Test is logically validated by showing the test
measures the behavior of interest.
Similar to logical validity approach with norm-
referenced validity.
© McGraw-Hill Higher Education. All rights reserved.
Decision Validity orDecision Validity orDecision AccuracyDecision Accuracy
Accuracy of a test in classifying subjects as proficient or nonproficient.
© McGraw-Hill Higher Education. All rights reserved.
Decision ValidityDecision Validity
Must be able to classify subjects as either true
masters or true nonmasters.This is often difficult to do.
A contingency table is then constructed.
© McGraw-Hill Higher Education. All rights reserved.
Decision ValidityDecision ValidityContingency TableContingency Table
Proficient NonproficientTrue Classification
Proficient
Nonproficient
Test Classification
A
D
B
C
© McGraw-Hill Higher Education. All rights reserved.
Validity Coefficient (or Validity Coefficient (or contingency coefficient)contingency coefficient)
C = (A + D) ÷ (A + B + C + D)
© McGraw-Hill Higher Education. All rights reserved.
Contingency TableContingency TableExampleExample
Proficient Nonproficient
Proficient
Nonproficient
Test Classification
40
3
6
12
True Classification
© McGraw-Hill Higher Education. All rights reserved.
Criterion-referencedCriterion-referencedValidity CoefficientValidity Coefficient
C = (A + D) ÷ (A + B + C + D)
C = (40 + 3) ÷ (40 + 6 + 12 + 3)
C = 43 ÷ 61
C = .70
© McGraw-Hill Higher Education. All rights reserved.
Interpretation of Validity Interpretation of Validity Coefficient “C”Coefficient “C”
If C = .50, classification was no better than
chance.
When C = .50, validity is low.
Values of C > .80 are desirable, but there are no
set rules for evaluating size of C.
© McGraw-Hill Higher Education. All rights reserved.
For Criterion-referenced Test For Criterion-referenced Test Validity:Validity:
Should have evidence of both
– domain-referenced validity and
– decision validity.
© McGraw-Hill Higher Education. All rights reserved.
Phi CoefficientPhi Coefficient
Another estimate of CRT validity. Correlation between two dichotomous
variables. Range from -1.0 to +1.0
© McGraw-Hill Higher Education. All rights reserved.
Phi CoefficientPhi CoefficientPhi = [(A x D) - (B x C)] .
(A + B)(C + D)(A + C)(B + D)
Proficient NonproficientTrue Classification
Proficient
Nonproficient
Test Classification
A
D
B
C
© McGraw-Hill Higher Education. All rights reserved.
Formative Evaluation of Chapter Formative Evaluation of Chapter ObjectivesObjectives
Define validity and outline the methods used to estimate it.
Describe the influence of test reliability on validity.
Identify those factors that influence validity.Select a valid criterion score based on
measurement theory.
© McGraw-Hill Higher Education. All rights reserved.
Chapter 4Chapter 4
ValidityValidity