Bayesian Estimation of a Multi-Unidimensional Graded-Response Item Response Theory Model

• In Event: 16.026 - Item Response Theory: Advances and Applications

Abstract

Graded response models (GRMs) have been developed for polytomous responses data. They have been widely used with instruments measuring one latent trait. However, in educational and psychological measurement, a cognitive or affective test often involves multiple subscales, where each polytomous item measures one trait. This paper introduced the item response model specific for this situation under the Bayesian framework. Monte Carlo simulations were conducted to evaluate the model performance and the results indicate that it performs well in various situations. In addition, increasing sample size, not test length, would result in improved accuracy in estimating model parameters. With the ability to model the intertrait relationship, the proposed model is more flexible and theoretically more appealing than the conventional GRM.

Authors

• Tzu Chun Kuo, Southern Illinois University - Carbondale

• Yanyan Sheng, Southern Illinois University - Carbondale