Evaluation of Limited Information Model Fit Statistics With Mixed-Format Tests

• In Event: Advances in Item Response Theory Models

Abstract

Many educational and psychological assessments contain multiple item formats. The purpose of this study is to examine the performance of limited information model-data fit statistics with mixed-format tests using simulated data. Specifically, we evaluate the Type I error rate and power level of limited information fit statistics, M2 and M2*, and their corresponding fit index RMSEA. Data are generated by manipulating different levels of sample sizes, number of items, proportions of polytomous items, and factor structures. The results from this simulation study are aimed to provide guidelines and suggestions for using limited information fit statistics with mixed-format tests.

Authors

• Mengyao Cui, Florida State University

• Zhongtian Lin, Florida State University

• Yanyun Yang, Florida State University