Değişen Madde Fonksiyonu Gösteren Ortak Maddelerin Test Eşitlemeye Etkisinin İncelenmesi

Abstract

In this study, it is aimed to examine the effect of anchor items with differential item functioning (DIF) on test equating. The effect of excluding anchor test items which showing DIF on equated scores and equating error was investigated by linear and nonlinear equating methods in non-equivalent groups with anchor test desing. DIF variables were determined as English, Spanish languages and the gender of the individuals who took the test in these languages. Mantel-Haenszel and SIBTEST techniques were used to identify items with DIF. Test equating was conducted using the Tucker, Levine, Chained equating method, which are linear equalization methods, Frequency estimation, Chained equipercentile and Circle-Arc Levine methods in nonlinear methods. The study data consists of the science literacy individual responses that have taken the 18th and 24th booklets of the PISA 2018 application in English and Spanish languages in computer based modules. According to the techniques, the effect of removing the items showing DIF at the B and C levels from the anchor test on the equating error was evaluated with the weighted mean square error (WMSE). As a result of the research, the exclusion of DIF items from the anchor test did not consistent changes in the equating errors. Inconsistency in equating errors; It was thought that the statistically significant difference between the averages of the booklets, the skill differentiation between subgroups of DIF variables and the degree of support the assumptions of the equating methods could be the source.