Development of mixed distribution item response models for the analysis of cross-classified multirater data and their application in teaching evaluation research (MixIRT-CCM)

点击翻译按钮获取中文摘要

The aim of the project is to extend measurement models of item response theory (IRT models) to the analysis of cross-classified multirater (CCM) data including latent classes (mixture components) or multiple measurement waves (longitudinal designs). The project builds on current research findings in the area of multimethod change measurement (e.g., Koch, 2013, Holtmann, 2017) and mixture IRT modeling for hierarchical data structures (e.g., Kelcey, McGinn, & Hill, 2014, Wang, 2011) and seeks to extend these new modeling approaches to so-called CCM measurement designs. The development of such models is of high scientific and practical interest, as CCM data are found in many areas of psychology and empirical educational research, particularly in the area of teaching evaluation research. The newly developed models are applied to teaching evaluation data in order to take a closer look at the convergent and discriminant validity of teaching evaluations at higher education institutions. In addition, the statistical performance of the models is investigated in Monte Carlo simulation studies.