A MATLAB Package for Markov Chain Monte Carlo with a Multi-Unidimensional IRT Model

摘要: Unidimensional item response theory (IRT) models are useful when each is designed to measure some facet of a unified latent trait. In practical applications, items not necessarily measuring the same underlying trait, and hence more general multi-unidimensional model should be considered. This paper provides requisite information description software that implements Gibbs sampler for such with two parameters normal ogive form. The developed written in MATLAB package IRTmu2no. flexible enough allow user choice simulate binary data multiple dimensions, set number total or burn-in iterations, specify starting values prior distributions parameters, check convergence Markov chain, as well obtain Bayesian fit statistics. Illustrative examples provided demonstrate validate use package.