An investigation of item bias for selection exams according to disability situations

In Turkey, there are exams known as Central Joint Exams for the eighth grade students in primary education. These exams are held twice a year for the eighth grade for six main courses. In this study, the selection test is conducted in mathematics and science. It was aimed to investigate whether the materials of have substance bias according to the disability status of the students with disabilities. For this purpose, Mantel-Haenszel (MH), Logistic Regression (LR) methods were used to determine whether the items of the selection exam contain differantial item functioning (DIF). The study group consistes of 1753 candidates for mathematics subtest and science subtest. The analyzes were conducted on 19 items for mathematics subtest and 20 items for science subtest. In the mathematics test, it was found that total of 2 items showed more than one comparison and 5 different items showed DIF in only one of the sub problems. In the science test, it was determined that 9 items had more than one comparison and 5 different items had DIF in only one of the sub-problems. It was decided whether or not the DIF-containing substances were biased by expert opinions. There was evidence of ıtem bias in 4 items in mathematics subtest and 8 items in science subtest.