Application of the Hierarchical Item Response Model to a Computer Adaptive Test of Graded Response Data

• In Event: Developments in Item Response Theory Modeling

Abstract

The hierarchical item response model is more flexible than the bi-factor or tri-level models as it allows for a mix of different level models as well as models with more than three level structures. When a computer adaptive test (CAT) is based on a hierarchical model, there are several item selection method choices. This study compared three item selection methods: the multidimensional item selection method, the local multidimensional item selection method, and the unidimensional item selection method, under different test length conditions. The result indicated that the multidimensional and local multidimensional item selection method provided the accurate general theta dimension estimation and the unidimensional item selection method provided accurate specific theta dimensional estimation when test length is medium.

Authors

• Wenhao Wang, The University of Kansas

• Neal M. Kingston, The University of Kansas

• Michael F. Hock, The University of Kansas

• Gail C. Tiemann, The University of Kansas

• Marcia H. Davis, Johns Hopkins University

• Stephen M. Tonks, Northern Illinois University