Adaptive testing, information and the partial credit model: A simulation study

In adaptive testing each individual is tested with a set of items that is selected to match his/her estimated ability at the time of testing. In its most sophisticated form an adaptive test is interactively administered by a computer that scores the test and uses the individual's pattern of correct and incorrect responses to select new items from an item bank. In this procedure items are selected during the process of administering a test (rather than as a part of a predetermined sequence) so that the items administered to each individual are appropriate in difficulty for that individual. The result is a test that is matched to each individual's ability and the test is neither too easy nor too difficult. In this form adaptive testing requires application of both computer technology and latent trait theory. This report works on extending adaptive testing to include the Rating Scale Model (RSM) (Andrich, 1978) and Partial Credit Model (PCM) (Masters, 1982), two Rasch models that allow items to be scored in ordered categories. It focuses on the nature of the information functions for the RSM and the PCM and the likely implications for adaptive testing, and reports on some simulations with RSM and PCM item banks each with items scored in three response categories.