Performance of Item Exposure Control Methods in Computerized Adaptive Testing: Further Explorations
American Educational Research Association Annual Meeting,
This study examined the effectiveness of the Sympson and Hetter conditional procedure (SHC), a modification of the Sympson and Hetter (1985) algorithm, in controlling the exposure rates of items in a computerized adaptive testing (CAT) environment. The properties of the procedure were compared with those of the Davey and Parshall (1995) and the Stocking and Lewis (1995) (SLC) conditional multinomial procedures within the purview of estimating examinee's abilities. Each of the exposure control methods was incorporated into the item selection procedure and the adaptive testing progressed based on the CAT design established for this study. The advantages and disadvantages of these strategies were considered under four item pool sizes and two desired maximum exposure rates and were evaluated in light of test security, test overlap rates, utilization of the item pool, and conditional standard errors of measurement. Also, the issue of the appropriate conditional sample sizes in deriving the exposure control parameters was considered in the present study. Simulation results show no effect of using the four conditional sample sizes. The SHC produced the most satisfactory results in terms of item security and test overlap rates followed by the SLC method. Results also show that as long as the control for item exposure was not exercised, optimal items could be administered to almost every examinee under any of the four item pools. Findings of this study provide useful insights on how item pool sizes and maximum item exposure rates affect the performance of the exposure control methods. (Contains 3 tables, 13 figures, and 18 references.) (SLD)
Chang, S.W., Ansley, T.N. & Lin, S.H. (2000). Performance of Item Exposure Control Methods in Computerized Adaptive Testing: Further Explorations. Presented at American Educational Research Association Annual Meeting 2000.