The aim of this paper is to propose a discretization method which can be applied to measures obtained using the Rasch model or, more in general, a model belonging to the class of the IRT models. The motivation of this proposal lies in the fact that there are methodologies that work with discretized variables, one such example is the Bayesian Networks. The idea is to use the informations from the Rasch model in order to forecast the answer of a subject to a representative item, and this answer represents the category assigned to the subject in the categorized version of his/her latent trait. In order to verify the goodness of this proposal, the new discretized variable is compared with a global single-item measure, under the hypothesis that this item is a possible observed discretization of the latent variable.
A proposal of a discretization method applicable to Rasch measures
GOLIA, Silvia
2017-01-01
Abstract
The aim of this paper is to propose a discretization method which can be applied to measures obtained using the Rasch model or, more in general, a model belonging to the class of the IRT models. The motivation of this proposal lies in the fact that there are methodologies that work with discretized variables, one such example is the Bayesian Networks. The idea is to use the informations from the Rasch model in order to forecast the answer of a subject to a representative item, and this answer represents the category assigned to the subject in the categorized version of his/her latent trait. In order to verify the goodness of this proposal, the new discretized variable is compared with a global single-item measure, under the hypothesis that this item is a possible observed discretization of the latent variable.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
