Investigating the Efficiency of Polytomous Hierarchical Generalized Linear Models in Exploratory Differential Item Functioning Assessment

• In Event: 52.070 - Multilevel Modeling: Estimation, Model Selection, and Sample Size

Abstract

The hierarchical generalized linear model (HGLM) is potential for DIF assessment due to its flexibility to structured data and capability for simultaneously assessing multiple items. Applying HGLM to assess DIF has been well studied for dichotomous items; however, for polytomous items, there still lacks evidence. The study aims to investigating the efficiency of polytomous HGLM (PHGLM) in DIF assessment, a simulation study and an empirical study are conducted for three critical issues, which are about locating DIF-free items, the effect of impact, and the practical application. The results show that PHGLM with the DIF-Free-then-DIF strategy is adequate for DIF assessment in polytomous items. The outcomes can provide practical solution for the application of PHGLM to the exploratory DIF assessment.

Authors

• Jyun-Hong Chen, National Sun Yat-sen University

• Cheng-Te Chen, National Chung Cheng University

• Ching-Lin Shih, National Sun Yat-sen Universit