多肢選択式テストによる測定の精緻化の試み

概要

論文の詳細を見る
In attempt to improve the measurement precision by a multiple-choice test, two IRT multi-category models were applied to an English test and compared to a binary model. The multi-category models were graded response model and nominal response model. Each item of the test consisted of 5 alternatives, which were classified according to their correctness into 3 categories : "correct, " "nearly correct, " and "far from correct." It was found that the two multi-category models showed the almost equal effect on improving the measurement precision compared to 2-parameter logistic model. The improvement was most remarkable for subjects with low ability.
東京大学の論文
1994-01-31