A Simulation Study to Compare CAT Strategies for Cognitive Diagnosis

This paper demonstrates the performance of two possible CAT selection strategies for cognitive diagnosis. One is based on Shannon entropy and the other is based on Kullback-Leibler information. The performances of these two test construction methods are compared with random item selection. The cognitive diagnosis model used in this study is a simplified version of the Fusion model. Item banks are constructed for the purpose of simulation. The major result is that the Shannon entropy procedure outperforms the procedure based on Kullback-Leibler information in terms of correct classification rates. However, Kullback-Leibler has slightly smaller item exposure rates than the Shannon entropy procedure. This study shows that the Shannon entropy procedure is a promising CAT criterion, but modification might be required to control the exposure rate.