Advanced modeling of item response data through the item response theory (IRT) or item factor analysis frameworks is becoming increasingly popular. In the social and behavioral sciences, the underlying structure of tests/assessments is often multidimensional (i.e., more than 1 latent variable/construct is represented in the items). This review provides an introduction and overview of flexMIRT 3.5, a commercially available statistical software program for the analysis of item response data that was developed specifically for fitting such complex models. Three models are used as motivating examples: (a) a bifactor model, (b) a multidimensional IRT (MIRT) model with correlated factors and (c) a diagnostic classification model with testlet structure. Discussion regarding flexMIRT input syntax and output files is provided and the topics of model fit evaluation, limited-information fit statistics in particular, and specifics of more recently introduced estimation methods and models are also covered.
- Multidimensional IRT
- diagnostic classification model
- limited-information fit statistics