Multilevel Item Response Theory Models

Abstract

The item response data structure is hierarchical since item responses are nested within respondents. Often respondents are also grouped into larger units and variables are available that characterize the respondents and the higher-level units. An item response modeling framework is discussed that includes a multilevel population model for the respondents and takes such a hierarchical data structure into account. An important application area is in education, where response observations are grouped in students and students grouped in schools. Several school e_ectiveness research studies are discussed. The hierarchical item response model is extended in several directions to handle latent explanatory variables, model latent individual growth, and identify clusters of respondents.