Gradual Maximum Information Ratio Approach to Item Selection in Computerized Adaptive Testing

June 29, 2009


For long-term quality control of computerized adaptive test (CAT) programs, optimizing the usage of the item bank (i.e., controlling item exposure rate) is critical. Selecting the best item often conflicts with the procedure for item exposure control, however. In this study, a newly proposed approach to item selection in CAT, in which the efficiency of items is considered in the early stages of CAT administration, was compared with the partial randomization method and the traditional maximized Fisher information (MFI) method. The simulation study found that the new approach greatly improved item bank utilization, compared with the other methods, while minimizing the compromise of test precision. The findings of this study could help practitioners find the best possible balance between test information and item bank utilization.