ValidityLecture Overview
• Overview of the concept• Different types of validity• Threats to validity and strategies for
handling them• Examples of validity issues from the
literature• Discussion of validity issues with respect to
student projects
Validity Descriptors
• “Hypothesis” validity
• “Construct” validity
• “Content” validity
• “Convergent” validity
• “Ecological” validity
• “Internal” validity
• “Statistical conclusion”validity
• “Concurrent” validity
• “External” validity
• “Predictive” validity
• “Criterion-related” validity
• “Discriminant” validity
Taxonomy of Validity
• Validity as it pertains to assessment
• Validity as it pertains to causal inference
• Validity as it pertains to generalization of findings to real-world phenomena
Validity Issues Surrounding Assessment/Measurement
Validity of an Assessment Tool
• Validity represents an overall judgment of the degree to which both empirical evidence and theoretical considerations support the interpretation of the score and the implications for action that this interpretation entails (Cronbach, 1971).
Validity of an Assessment Tool
• Score validation is an empirical evaluation of the meaning and consequences of measurement (Messick, 1989).
Validity Features
• Validity applies to all assessments, including performance/behavioral assessments
• Validity is not a property of the “test” per se, but rather of the meaning of the test score
• Validation is an ongoing process
Features
• Validity is not just a measurement principle, it is a social value that has powerful implications whenever evaluative judgments and decisions are made
Types of “Assessment” Validity
• Content validity– Degree to which “Test” items adequately
sample the universe of relevant items for a given domain
Types of “Assessment” Validity
• Criterion-related validity– Degree to which a “Test” score relates to some
relevant external criterion• Concurrent validity
• Predictive validity
Types of “Assessment” Validity
• Construct validity*– Ongoing, integrated summary of the evidence
supporting the interpretation and utility of “Test” scores
– Combines information from content validity, criterion-related validity, and discriminant/convergent validity
Convergent/Discriminant Construct Validation
• Convergent validity– Empirical evidence demonstrating communality
between the test score and other indicators of the same construct
• Discriminant validity– Empirical evidence demonstrating a lack of
communality with the test score and indicators of a different construct
Threats to Assessment Validity
• Construct underrepresentation– Exists when the assessment fails to include
important facets of the construct (i.e., assessment is too narrow)
– Examples?
INSTRUCTIONS: Circle the one phrase that best represents the extent to which you agree with the item. If any of the items concern something that is not part of your experience (e.g., “it scares me when I feel shaky” for someone who has never trembled or had the “shakes”), answer on the basis of how you think you might feel if you had such an experience. Otherwise, answer all items on the basis of your own experience. 1. It is important to me not to appear nervous. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 2. When I cannot keep my mind on a task, I worry that I might be going crazy. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 3. It scares me when I feel 'shaky' (trembling). VERY LITTLE A LITTLE SOME MUCH VERY MUCH 4. It scares me when I feel faint. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 5. It is important to me to stay in control of my emotions. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 6. It scares me when my heart beats rapidly. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 7. It embarrasses me when my stomach growls. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 8. It scares me when I am nauseous. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 9. When I notice that my heart is beating rapidly, I worry that I might have a heart attack. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 10. It scares me when I become short of breath. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 11. When my stomach is upset, I worry that I might be seriously ill. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 12. It scares me when I am unable to keep my mind on a task. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 13. Other people notice when I feel shaky. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 14. Unusual body sensations scare me. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 15. When I am nervous, I worry that I might be mentally ill. VERY LITTLE A LITTLE SOME MUCH VERY MUCH 16. It scares me when I am nervous. VERY LITTLE A LITTLE SOME MUCH VERY MUCH
Threats to Assessment Validity
• Construct-irrelevant variance– Exists when the assessment contains reliable
variance associated with other distinct constructs (i.e., assessment is too broad)
– Examples?
Evidence for Assessment Validity
• Evidence of content relevance and representativeness
• The extent to which test scores are consistent with theoretical predictions
• Evidence examining the extent to which score properties and interpretations generalize to and across groups, settings, and tasks
Evidence for Assessment Validity
• Evidence on the fidelity of the scoring structure to the structure of the construct being tapped
• Evidence from criterion-related studies including convergent and discriminant studies
• Evidence pertaining to the consequential aspect of test use and score interpretation, especially as it relates to issues of bias, and fairness
Strategies for Enhancing Assessment Validity
• Avoid sole reliance on measures that lack validation data (e.g., new author-constructed measures)
• Employ multiple indicators of the focal construct whenever possible
• Employ indicators from more than one assessment modality domain
• Discussion?
Drawing Valid Inferences about Causal Relationships
Types of Validity Pertinent to Drawing Causal Inferences
• Internal validity– Degree to which causal inferences can be made
between a measured or manipulated variable (i.e. independent variable) and another measured variable (dependent variable)
Types of Validity Pertinent to Drawing Causal Inferences
• Statistical conclusion validity– Concerned with sources of random error and with the
appropriate use of statistics and statistical tests (as opposed to systematic bias as in the case of internal validity)
Types of Validity Pertinent to Drawing Causal Inferences
• External validity– Refers to the degree to which the observed
causal relationship is generalizable across persons, settings, and occasions
– Important distinction between generalizing to a specified population (or setting or occasion) vs. generalizing across populations
Types of Validity Pertinent to Drawing Causal Inferences
• Construct validity*– The degree to which causal inferences
concerning one variable’s effect on another can be generalized to examplars of the constructs in question
– In every day practice, this form of validity deals with the issue of “confounds”
Threats to Internal Validity
• History
• Maturation
• Testing
• Instrumentation
• Statistical regression
Hypothetical Example of Maturation Threat
05
101520253035404550
Jan Feb March April
ProzacLuvox
Hypothetical Example of Maturation Threat
05
101520253035404550
Jan Feb March April
ProzacLuvox
Hypothetical Example of Maturation Threat
10
15
20
25
30
35
40
45
Jan Feb March April
ProzacLuvoxPlacebo
Example of a“Testing” Threat
0
10
20
30
40
50
60
% w
ho P
an
ick
ed
CBT EST WL
PrePost
Data from Jaimez & Telch (in preparation)
Threats to Internal Validity
• Selection
• Mortality
• Ambiguity about the direction of causal influence
Threats to Internal Validity
• Diffusion of treatments
• Compensatory equalization of treatments
• Compensatory rivalry by respondents
• Resentful demoralization of respondents
Threats to Internal Validity
• Compensatory equalization of treatments
• Compensatory rivalry among participants
• Resentful demoralization
• Mortality
Threats to Construct Validity About Cause and Effect
• Construct underrepresentation– Mono-operation bias– Mono-method bias– Confounding constructs and levels of
constructs
Threats to Construct Validity About Cause and Effect
• Construct irrelevancies (i.e., confounds)– Interaction of different treatments– Hypothesis-guessing within experimental
conditions– Evaluation apprehension (demand
characteristics)– Experimenter expectancies– Interaction of testing and treatment
Threats to External Validity
• Interaction of selection and treatment
• Interaction of setting and treatment
• Interaction of history and treatment
Strategies for Enhancing External Validity
• Employ random sampling to obtain a representative sample if time, resources, and feasibility permit
• Employ heterogeneous samples whenever possible
• Conduct analyses to determine whether the causal relationship holds across characteristics of subjects, settings, etc