The Effect of Outliers Treatment Method on the Effectiveness of Equating Two Test Forms

Abstract

This study aimed at revealing the effectiveness of equating two test forms according to item response theory using different outlier treatment methods. To achieve that, the linear equating method was used to equate two equivalent test forms of the National Test to Control the Quality of Education in Mathematics for fourth grade, based on the ability (θ) according to random groups design in equating.  The first form consisted of (35) multiple choice items and the second form consisted of (36) multiple choice items. Responses of (2000) students were randomly selected for each form. Standard error of equating and cross-validation were used for judging equating effectiveness. The results showed that replacement of outliers lead to an increase in tests equating effectiveness. Also, there are significant differences between the means of standard errors for the equated abilities, in favor of replacing the outliers at all ability levels compared with either keeping or removing the outliers. It also revealed that effectiveness of equating is better whenever the ability level tends toward the center of ability distribution