The Performance of Item Response Theory Model Selection Methods With Mixed-Format Tests

• In Event: Item Response Theory: Applications and Extensions

Abstract

Item response theory (IRT) model selection studies have been conducted with single-format tests. This simulation study examined the performance of model selection methods used to select among a set of mixed-format IRT models. Data were generated according to a 3PL/GPC or 1PL/PC combined IRT model. Data were subsequently fitted to these two models and a 2PL/GPC model to examine the accuracy of five information-based model selection criteria when selecting among the three models. These criteria included the AIC, AICC, BIC, HQIC, and CAIC. The model selection criteria examined did not accurately select the 3PL/GPC model whereas they did accurately select the 1PL/PC model under the conditions manipulated. Differences among the performance of the criteria under certain conditions are discussed.

Authors

• Tiffany Ann Whittaker, The University of Texas - Austin

• Wanchen Chang, The University of Texas - Austin

• Barbara G. Dodd, The University of Texas - Austin