Assessing the Effects of Different Item Parameter Profiles in Mixture Rasch Models

• In Event: Issues of Rasch Dimensionality, Scaling, and Fit

Abstract

The allowance for the different item parameter profiles across latent classes is one of the main features of mixture Rasch models. The effects of different item parameter profiles across latent classes on the model performance, however, have not been extensively studied. In this study, the item difficulty profiles are systematically varied based on three properties identified in previous studies to represent a wide range of possible patterns for item difficulties across latent classes. The effects of the differences in item difficulty profiles are assessed on the accuracy of a) item parameter recovery, b) person parameter recovery, and c) correct model identification. In addition, the impacts of test length, sample size, and class proportion are investigated.

Authors

• Youngmi Cho, University of Maryland

• Hong Jiao, University of Maryland

• George B. Macready, University of Maryland