Characterizing the Latent Classes in a Mixture IRT Model Using DIF

Mixture IRT models address the heterogeneity in a population by extracting latent classes and allowing item parameters to vary between latent classes. Once the latent classes are extracted, they need to be further examined to be characterized. Some approaches have been adopted in the literature for this purpose. These approaches examine either the examinee or the item characteristics conceptually or statistically. In this study, we propose a two-step procedure for characterizing the latent classes. First, a DIF analysis can be conducted to determine the items that function differentially between the latent classes using the latent class membership information. Then, the characteristics of the items with DIF can be further examined to use this information for characterizing the latent classes. We provided an empirical example to illustrate this procedure.