Investigation of item bias for English subtest of TEOG

In this study, differential functioning items according to gender and school type (private vs public) in English subtest of TEOG in 2015 – 2016 Academic Year Spring Semester were investigated. Expert opinions were sought for the items, that were found as showing differential item functioning, in order to determine whether there was an item bias. Mantel-Haenszel (MH), Logistic Regression (LR) and SIBTEST methods were used in the study of the Differential Item Functioning (DMF) analysis. The data includes the answers in 20 items English test given by 9108 8th grade students randomly selected from total 1133857 students. Before the analysis, the unidimensionality of the English subtest was examined. Then, descriptive statistics, reliability analyzes and item statistics were calculated for each subtest created according to gender and type of school. In the DMF analysis, firstly gender-based analyzes were carried out and it was revealed that items 5, 9, 11 and 16 differed at B (medium) level. Among these items, item 5 showed DIF in only SIBTEST, item 9 and 11 showed DIF in both MH and SIBTEST, and item 16 showed DIF in only MH. Based on expert opinions, it was decided that only item 11 might show bias in favor of male students. According to the analysis made based on school type variable, only item 4 was determined as showing B level DMF in both MH and SIBTEST methods. Experts opinions on this item indicated that this item might show bias in favor of private school students. Keywords: TEOG, differential item functioning, item bias, Mantel Haenszel method, Lojistik Regresyon method, SIBTEST method