Professor Carol Woods
 
Department of Psychology, Washington University
 
Title: Empirical Selection of Anchor Items for IRT-LR-DIF
 

Abstract: IRT-LR-DIF is a method used to evaluate the degree to which test or questionnaire items are biased towards one group of people compared to another group. Groups may be defined by gender, ethnicity, etc. IRT-LR-DIF requires that a few items on the test have the same measurement properties in both groups -- these are called anchor items. Many complicated and untested procedures for empirically identifying anchor items are suggested in the literature. In this talk, IRT-LR-DIF will be introduced, and a simple method for empirically identifying anchor items that performs well in simulations will be suggested.