Several methods have been developed for use on constrained adaptive testing. Item pool partitioning, multistage testing, and testlet-based adaptive testing are methods that perform well for specific cases of adaptive testing. The weighted deviation model and the Shadow Test approach can be more generally applied. These methods are based on different ideas about how to optimize the performance of computerized adaptive testing (CAT), and they have different advantages. In this paper, both ideas are combined, and a new approach for selecting items in CAT, the Extended Shadow Test approach is presented. The performances of the Extended Shadow Test Approach, the Shadow Test approach, and the weighted deviation model are compared using simulation and existing item pools. The Extended Shadow Test approach shows accurate performance. Recommendations are given about its use.
|Name||OMD research report|
|Publisher||University of Twente, Faculty of Educational Science and Technology|