Polytomous Item Explanatory Item Response Theory Models With Random Item Errors: Simulation and Empirical Studies

• In Event: Innovative Approaches to Item Response Theory

Abstract

Despite the methodological usefulness of item explanatory IRT models, their applications have been considerably less investigated for polytomous data than dichotomous data. Recently, polytomous item explanatory models with random item errors were proposed building on Many-Facet Rasch Model (MFRM) and Linear Partial Credit Model (LPCM), considering the uncertainty in explanation and/or random nature of item parameters. This paper aims to conduct simulation and empirical studies using a Bayesian method to demonstrate practical applications and implications of the proposed models. The results show that the proposed models with random item errors performed better than the models without random item errors and demonstrated practical differences in interpreting the item property effects between item location explanatory MFRM and step difficulty explanatory LPCM approaches.

Authors

• JinHo Kim, University of Seoul

• Mark R. Wilson, University of California - Berkeley

• Sophia Rabe-Hesketh, University of California - Berkeley