Accounting for Low Reliability in Multilevel Differential Item Functioning Analysis: A Latent Variable Approach

• In Event: 65.080 - Roundtable Session 24
  In Roundtable Session: 65.080-18 - Reliability: Methods and Application

Abstract

The purpose of the current study is to examine the comparative performance of logistic regression (LR), hierarchical logistic regression (HLR), and hierarchical logistic regression with a latent covariate (HLR-LC) for differential item functioning (DIF) detection. The results of the current study will provide evidence to support the implementation of HLR-LC to account for low reliability of short tests in DIF analysis when the primary sampling units are clusters.

Authors

• Ying Jin, Middle Tennessee State University

• Christian Flack, Middle Tennessee State University