Multilevel Versus Simple Modeling for Differential Item Functioning Detection

• In Event: Multilevel Modeling: Estimation, Model Selection, and Sample Size

Abstract

The purpose of the study was to examine the comparative performance of multilevel versus simple modeling for differential item functioning (DIF) detection under a condition that has not been studied previously, but may be encountered rather frequently in practice. The performance of two simple DIF methods, the Mantel-Haenszel test (MH) and Logistic Regression (LR), was compared to the performance of a complex DIF method Hierarchical Logistic Regression (HLR). Simulation indicated that HLR, LR, and MH performed equivalently in terms of significance tests under most conditions. Effect size estimates of HLR, LR, and MH were equally accurate and consistent under the Rasch model, and were less accurate and consistent under the two parameter item response theory model.

Authors

• Ying Jin, Middle Tennessee State University

• Nicholas D. Myers, University of Miami

• Soyeon Ahn, University of Miami