Modeling differential item functioning with group-specific item parameters: A computerized adaptive testing application

Many important decisions are made based on the results of tests administered under different conditions in the fields of educational and psychological testing. Inaccurate inferences are often made if the property of measurement invariance (MI) is not assessed across these conditions. The importance of MI is even greater when test respondents are compared based on their responses to different items, such as the case in computerized adaptive testing (CAT), because the existence of items that exhibit differential item functioning (DIF) can produce bias within a group as well as between groups. This article demonstrates a straightforward psychometric method for conducting a test of measurement invariance (MI) and illustrates a method for modeling DIF by assigning group-specific item parameters in the framework of IRT. The article exemplifies two applications of the method for a CAT used in a high stakes international organizational assessment context. These examples pertain to context effects due to the test administration method (computer based linear test vs. CAT), and the context effects due to language in a CAT