Analysis of Differential Item Functioning of PISA 2015 Mathematics Subtest Subject to Gender and Statistical Regions

Abstract

It is accepted that the presence of Differential Item Functioning (DIF) in large scale examinations may be an indication of bias. The aim of the present study was to analyze whether dichotomous (1-0) items in the PISA 2015 mathematics subtest exhibit DIF with regard to gender and statistical regions. The study was carried out using the data of 2409 students who took part in PISA 2015 and answered all mathematics questions. Rasch model was used via Winsteps software to determine whether the items exhibit DIF or not. Confirmatory Factor Analysis (CFA) was applied to the sixty-three mathematics questions in the clusters. The modification indices and goodness of fit values were examined for CFA and a total of eight items that disrupt the model structure were excluded from the test. DIF analyses were carried out for the 55 items that were observed to be one-dimensional. The analysis results based on gender indicated that five items exhibit DIF. Two of these items exhibit DIF in favor of girls, while three in favor of men. Statistically significant DIF findings were observed in all items when the analyses results based on statistical regional units were analyzed. While at least 10 DIF cases were observed as a result of the binary territory comparison on an item basis, maximum 38 DIF cases were observed. Minimum DIF was observed in item Q4 as a result of regional comparisons, whereas maximum DIF was observed in items Q47 and Q50.