Determining if New Psychological Tests Have Construct Validity
Validity
Define Construct Validity and Explain and Illustrate three ways in which a researcher can determine whether a new psychological tests has construct validity. In addition briefly explain what is meant if a test is biased in terms of construct validity.
Construct Validity
•
•
•
•
•
•
•
•
•
Construct Validity: extent to which a measure measures theoretical construct or trait
Developmental changes:
▪ when tests should show change with age
▪ age differentiation in intelligence in childhood
▪ necessary but not sufficient
▪ culture affects development
Correlation with other measures
▪ measure same construct
▪ moderately high
▪ Incremental validity: stronger correlation between new measure and construct
Convergent and Discriminant validity:
▪ comparison with external tests/constructs/criteria
▪ high correlation between variable with which it should correlate; minimal correlation with variables with which there should be difference
Factor validity
▪ factor analysis – to analyse interrelationships of variables
▪ determine underlying factors/constructs that are tapped by a measure
• simply description of construct, composition of test
• do these correlate with what is supposed to be measured
▪ factorial validity: correlating test with itself
▪ Uses: new measures; new contexts
Internal Consistency
▪ correlation of scores on each item or sub test with total score
▪ items should show discriminatory power
• remove those that don't
▪ criterion is the test itself
▪ not sufficient – need to consult external criteria
Experimental Interventions
▪ additional source
▪ pre-test, intervention, post-test
▪ score lower on pre-test
• when items don't show this, they may be too easy, too difficult, be invalid for construct ▪ eg: depression
Differential validity
▪ differentiate between characteristics of individual's or groups or organisations
▪ eg: personality measures yielding different personality profiles
Test Bias
Define Construct Validity and Explain and Illustrate three ways in which a researcher can determine whether a new psychological tests has construct validity. In addition briefly explain what is meant if a test is biased in terms of construct validity.
Construct Validity
•
•
•
•
•
•
•
•
•
Construct Validity: extent to which a measure measures theoretical construct or trait
Developmental changes:
▪ when tests should show change with age
▪ age differentiation in intelligence in childhood
▪ necessary but not sufficient
▪ culture affects development
Correlation with other measures
▪ measure same construct
▪ moderately high
▪ Incremental validity: stronger correlation between new measure and construct
Convergent and Discriminant validity:
▪ comparison with external tests/constructs/criteria
▪ high correlation between variable with which it should correlate; minimal correlation with variables with which there should be difference
Factor validity
▪ factor analysis – to analyse interrelationships of variables
▪ determine underlying factors/constructs that are tapped by a measure
• simply description of construct, composition of test
• do these correlate with what is supposed to be measured
▪ factorial validity: correlating test with itself
▪ Uses: new measures; new contexts
Internal Consistency
▪ correlation of scores on each item or sub test with total score
▪ items should show discriminatory power
• remove those that don't
▪ criterion is the test itself
▪ not sufficient – need to consult external criteria
Experimental Interventions
▪ additional source
▪ pre-test, intervention, post-test
▪ score lower on pre-test
• when items don't show this, they may be too easy, too difficult, be invalid for construct ▪ eg: depression
Differential validity
▪ differentiate between characteristics of individual's or groups or organisations
▪ eg: personality measures yielding different personality profiles
Test Bias