Comparison of Model-Data Fit Evaluation Methods for Multidimensional Item Response Theory and Structural Equation Modeling: A Monte Carlo Study

• In Event: 70.066 - Roundtable Session 28
  In Roundtable Session: 70.066-6 - Applications of Advanced Item Response Theory Models

Abstract

Item response theory (IRT) and structural equating modeling (SEM) are two families of techniques to model the relationship between observed responses and the latent variables. Given the similarities between SEM and IRT, it is necessary to further examine the performance of SEM- and IRT-related fit statistics/indices in one study, which serves the purpose of this study. Specifically, we evaluate SEM-related fit statistics/indices when data are generated based on both SEM and Multidimensional IRT (MIRT) models; we also evaluate MIRT-related fit statistics when the data are generated based on both SEM and MIRT models. Data are generated by manipulating different levels of sample sizes, model sizes, number of scale points, and non-normality of ability distribution.

Authors

• Mengyao Cui, Florida State University

• Yanyun Yang, Florida State University