Multidimensional Random Coefficients Multinomial Logit Differential Item Functioning (DIF) Decomposition Modeling for a Testlet Item DIF Investigation

• In Event: Comparing, Decomposing, and Exploring Models

Abstract

DIF for a testlet item may be decomposed into item specific differential functioning (IS_DF) and testlet specific differential functioning (TS_DF), i.e., DIF = IS_DF + TS_DF. Previously suggested DIF decomposition models were based on the multilevel modeling framework and estimated by the residual pseudo likelihood method. This study proposes an alternative Rasch DIF decomposition modeling approach which is based on the multidimensional RCML model and estimated by the MMLE-EM algorithm, which make the model estimation more efficient. Also, this study includes DIF parameter testing results and the evaluation of the proposed model’s capacity to cope with the impact (group ability difference on the garget dimension), which were not investigated in the previous studies.

Authors

• Insu Paek, Florida State University

• Hirotaka Fukuhara, Pearson