Polytomous Item Explanatory Item Response Theory Models: Application to Carbon Cycle Assessment Data

• In Event: Modern Applications in Item Response Theory

Abstract

Despite methodological usefulness of item explanatory IRT models, their applications have been considerably less discussed for polytomous items than for dichotomous items. It is mainly due to the difficulty of item parameterization using item properties in a statistical model. To investigate polytomous item explanatory models, the partial credit model is extended in two ways under the multivariate generalized linear mixed modeling framework. One is that item location parameters are explained by item property effects, and the other is that step difficulty parameters are decomposed into weighted sums of item property effects. The two proposed models worked differently for Carbon Cycle assessment data, based on types of incorporated item property effects and their target parameters to be explained.

Authors

• Jinho Kim, University of California - Berkeley

• Mark R. Wilson, University of California - Berkeley