Investigating Latent Class Characteristics With Covariates by Using Mixture Rasch Model

• In Event: Item Response Theory With Complex Data

Abstract

Mixture IRT models with covariate are useful for studying examinee characteristics that affect examinee classification into latent classes. In this way, it permits us to better understand examinee performance differences in different classes. To investigate the degree to which more detailed information may be garnered, we analyzed the fraction computation data using a mixture Rasch model(MRM; Rost,1990) to detect the classes, then compared that with a MRM with covariate to better explain the reasons why the latent groups might have formed. The results suggested that including a covariate (math disability) in the model affected mixing proportions and class membership. This covariate provided additional information that helped classify examinees into different classes and helped interpret examinees’ error patterns within each class.

Authors

• Selay Zor, University of Georgia

• Allan S. Cohen, University of Georgia

• Brian A. Bottge, University of Kentucky

• Linda Joy Gassaway, University of Kentucky