Cognitive Diagnostic Models and Computerized Adaptive Testing: Two New Item-Selection Methods That Incorporate Response Times

Abstract

A recent paper proposed an item-selection approach for computerized adaptive testing (CAT) in which the psychometric information per time unit is maximized. The current research extended this methodology to adaptive tests combined with use of a cognitive diagnostic model (CDM). Two new item-selection methods are introduced for the combination of CDMs and CAT: posterior-weighted Kullback-Leibler information per-time-unit, and mutual information per-time-unit. These methods were compared with standard procedures in which the amount of time required to complete an item is not considered. Simulation conditions with and without attribute-balancing constraints indicated that, on average, the new methods required more items but took less time than the standard procedures, while achieving comparable classification accuracy.