A multidimensional Item Response Theory approach for the University guidance

The aim of this work is to investigate a multidimensional item response theory approach applied to a competence test for university guidance. The possibility of the simultaneous estimation of item parameters for two sections of the guidance test is explored by using the multidimensional two-parameter normal ogive model. Two different estimation methods are considered: the full-information factor analysis and the Gibbs sampler algorithm in the Markov chain Monte Carlo framework. Furthermore, the incomplete design is taken into account. In particular, advantages of using the Gibbs sampler respect to the ML estimation are discussed.