가교문항-고정 다집단 IRT 추정을 위한 BILOG-MG 프로그램의 적용 가능성 진단

초록

Most testing programs have item pools that consist of items from many alternate forms of a test. When analyzed by item response theory (IRT), all items in an item pool are expected to have their item parameters placed on a common IRT ability scale. When two or more test forms are developed to include some items in the pool as an anchor under the common-item nonequivalent groups equating design (in which the anchor items need not be the common items between test forms), multiple-group and multiple-form test data are obtained. The purpose of this study is to diagnose the applicability of the computer program BILOG-MG for fixed-anchor multiple-group IRT parameter estimation for such multiple-group and multiple-form test data. For this purpose, two BILOG-MG based fixed-anchor estimation methods, Simple Update and DeMars-Jurich, proposed for the analysis of single-group test data were extended to the ones for fixed-anchor multiple-group IRT estimation, and their performances were examined through simulations of both single-group and multiple-group testing. The simulation results showed that both of the Simple Update and DeMars-Jurich fixed-anchor IRT estimation methods worked well with the single-group test data, but the two methods did not properly recover the true parameters of non-anchor items with the multiple-group test data. The results suggest that when using BILOG-MG to conduct fixed-anchor IRT parameter estimation with multiple-group data, the Simple Update or DeMars-Jurich method should be applied to each group’s data.

키워드