同時尺度調整法による垂直的等化の検討

概要

論文の詳細を見る
Using the concurrent calibration in the item response theory, this research examines the vertical equating of the average of two populations through simulation. Simulated data were generated using 2-parameter logistic model with the vertical equating on the assumption of conditions in which an individual test was conducted for two populations having a significantly different ability (anchor items included in both tests). The population average and variance were estimated by the Mislevy's (1984) method and the method of averaging the maximum likelihood estimates of subjects' abilities, and compared the results of both methods. The following explains the simulation results. The Mislevy's method showed the better results than the other method. For the reproduction of difference between the averages of two populations under conditions in which concurrent calibration was used, it was also found that the results would not become better proportionally to the number of test items increased. This finding disproves the general concept that increasing the number of test items will bring better results. Another finding is that estimated results improve proportionally to the number of anchor items between the two tests increased.
文教大学の論文
1998-12-20