Summary: | 碩士 === 輔仁大學 === 統計資訊學系應用統計碩士班 === 103 === In this study, we apply sequential one-side confidence interval estimation procedures with β– protection to adaptive multicategory testing. The results indicated thatthere were no significant differences in average length of test for every ability interval.There are no indifference region with varying wide of confidence, it can reduce misclassification rate.With the truncation of test, both misclassification rate and length of test can be reduced. For the item selectionstrategies, FI can reduce length of test in SIE.However, KL is more efficient than FIin SPRT.
|