A New Tool for Fitting Polytomous Item Response Theory Models

• In Event: Roundtable Session 117
  In Roundtable Session: Discussions in Item Response Theory

Abstract

In this paper I establish the models for the manifest probability in polytomous item response theory (IRT) models, and use the obtained log-linear models to develop new tools for analyzing the polytomous item response data. By fitting the derived log-linear model for the manifest probability, the solutions for the original polytomous IRT model are also obtained. I will focus on the family of partial credit models (PCMs). Concomitant information such as item properties and person properties can also be incorporated in the models as covariates. The effectiveness of the developed models and the pseudolikelihood estimation method is demonstrated by a series of simulation studies and the application to a survey on verbally aggressive behavior.

Author

• Zhushan Mandy Li, Boston College