Assessing model data fit of unidimensional item response theory models in simulated data

Abstract

The pupose of this paper is to give an example of how to assess the model-data fit of unidimensinal IRT models in simulated data. Also, the present research aims to explain the importance of fit and the consequences of misfit by using simulated data sets. Responses of 1000 examinees to a dichotomously scoring 20 item test were simulated with 25 replications. Also, data were simulated to fit the 2-PL model. 4-step procedure has been used for model-data fit and BILOG was used as software. Results were discussed in the frame of the literature.