On the extended specific objectivity of some pseudo–rasch models applied in testing scenarios

Abstract

The design of a Computer Adaptive Testing (CAT) system assumes the existence of an item pool containing properly calibrated items. The calibration is based on an Item Characteristic Curve (ICC). In this paper two mathematical ICC models, and how these models properly fit into the concept of extended Rasch specific objectivity, are under analysis. The results make clear that the comparison between two items depends on subdomains of the complete domain of the corresponding ICC’s. The introduced models are also useful to describe the characteristics of skewness and bimodality in the population, where classical models commonly fail.