Applying the Testlet Response Theory to the Random Facet px(i:c) Design of Generalizability Theory

• In Event: Modeling Frameworks for Measurement Issues

Abstract

Briggs and Wilson (2004, 2007) GIRM approach for estimating variance components in Generalizability Theory (GT) and demonstrated this approach using the Rasch model under the single random facet pxi design. Rather than using observed scores as the traditional GT approach does, the GIRM approach uses expected scores, which are obtained by employing the Markov Chain Monte Carlo (MCMC) technique. This study further investigates GIRM and extends their work from a single facet design to a random facet px(i:h) design using real data, where h represents testlet or context. Under this design, a multidimensional Testlet Response Theory (TRT; Wainer & Wang, 2000) model is used to estimate the expected scores within MCMC.

Authors

• Yuehmei Chien, Pearson

• Robert L. Brennan, University of Iowa