New Article in the Journal of Computerized Adaptive Testing (JCAT)
In recent years, several methods have been proposed to detect compromised items (CI) and examinees with preknowledge (EWP) simultaneously. However, most of these methods are limited in one of two ways: (1) The method was specifically designed for non-adaptive tests and might therefore not be suitable for adaptive tests, or (2) the method involves the analysis of more than just the item scores. In this paper, a new method is developed for computerized adaptive tests that requires only an analysis of the item scores. The performance of the method is evaluated using simulations in which several factors are manipulated, such as test length, the percentage of CI, the percentage of EWP, and EWP ability distribution. Across most conditions, the method is shown to produce reasonable false positive rates and favorable true positive rates for both items and examinees.
