Performance of Non-parmametric Approches to Model Selection in Mixture Psychometric Models

• In Event: Bayesian Techniques

Abstract

A pseudo non-parametric method for model selection in mixture modeling applications has recently been proposed as an improved alternative to fully non-parametric methods. We evaluate performance of this recent method for mixture and hidden Markov IRT models. Additionally, we discuss its utility over traditional model selection approaches.

Authors

• Clifford Erhardt Hauenstein, Johns Hopkins School of Medicine

• Eunbee Kim, GEORGIA INSTITUTE OF TECHNOLOG