Detecting Intersectional Differential Item Functioning: A Comparison of Two Methods

• In Event: Sunday Roundtable 9:45 am
  In Roundtable Session: Exploring Equity and Fairness in Educational Measurement (Table 39)

Abstract

Differential Item Functioning (DIF) analysis is a technique to detect whether certain items favor specific subgroup members after controlling their overall performance, and it has been instrumental in ensuring test fairness. While many DIF studies analyze individual demographic variables separately, there is a growing recognition of intersectional DIF analysis, which captures the potential compound DIF effect by multiple demographic variables. This study introduces two methods to probe intersectional DIF: the regularization method and the item-focused tree method. We use an empirical dataset to illustrate and compare these methods. Results reveal three items with intersectional uniform DIF detected by the regularization method and none by the tree method, while the tree method detected one item with intersectional non-uniform DIF.

Authors

• He Ren, University of Washington

• Chun Wang, University of Washington

• Min Li, University of Washington

• Miranda Parker, San Diego State University