Comparison of Item Targeting Strategies for Pass/Fail Computer Adaptive Tests.

The most useful method of item selection for making pass-fail decisions with a Computerized Adaptive Test (CAT) was studied. Medical technology students (n=86) took a computer adaptive test in which items were targeted to the ability of the examinee. The adaptive algorithm that selected items and estimated person measures used the Rasch model and a version of maximum likelihood estimation. The stopping rule was based on confidence in the pass/fail decision. Results indicate that when test length is sufficient, targeting items at the ability of the examinee and using a confidence level stopping rule results in the most efficient computer adaptive test for making a pass/fail decision. Examinees whose ability is clearly above or below the pass/fail point then take a minimum number of items, but those whose ability is near the pass point take a test of precision comparable to a test of items targeted at the pass/fail point. An appendix contains an examinee map for the test and a map key. (Contains two tables, three figures, and six references.) (SLD)