동등집단 설계를 사용하는 수직척도화에서 공통 능력척도의 개발을 위한 IRT 추정 방법의 비교

초록

In item response theory (IRT) based vertical scaling, a common ability scale can be developed using the ‘multiple-group concurrent estimation’ method or the ‘separate estimation and scale-linking’ method. This study investigated, through computer simulations, the comparative performances of these two IRT estimation methods under the equivalent groups design. In the equivalent groups design, three grades were involved and examinees in each grade were randomly assigned to take either the test level designed for their grade or the test level designed for one grade below their grade. The computer simulations showed that overall the separate estimation and scale-linking method performed better than the multiple-group concurrent estimation method in developing a vertical common scale. This comparative results are inconsistent with those from the previous research conducted under the common-item and scaling-test designs. The possible reasons that the separate estimation and scale-linking method could be superior to the multiple-group concurrent estimation were discussed.

키워드