SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Imprint / Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University

Contributions: Abstract

Bayesian IRT: A bivariate application and software validation

Samantha Cook Donald B. Rubin
Harvard University

Item response theory (IRT) is a method for analyzing test data in which the test items themselves are analyzed in addition to the test-takers' abilities. The underlying models account for the fact that test questions may not be equally difficult, and also produce results that are more generalizable and less test-dependent.

We present a Bayesian bivariate IRT model developed for a study of working memory impairments in schizophrenia patients. Using the Bayesian framework allows for straightforward extension of standard univariate IRT models to the bivariate case.

Bayesian modeling also allows for principled validation of the model-fitting software; we illustrate this validation strategy using the bivariate IRT model.