Analysis of PISA 2015 science literacy test in terms of differential item functioning

In this study, it was investigated whether the items included in the PISA 2015 Science Literacy test show Differential Item Function (DIF) in terms of gender, the residential area of the school and socioeconomic level. In DIF analysis, Mantel Haenszel (MH) and Logistic Regression (LR) methods were used and then the correlation between the results from these two methods were examined. The data of gender and socioeconomic level variables used in this study were obtained from the student's questionnaire while the data of the residential area of the school variable was obtained from the school questionnaire. The study was carried out with data from 170 students who answered 35 common questions in booklets 32, 41, 44, 53, 56, 65, 68, 77, 80, 89 and 92 in the PISA 2015 application. In order to determine the general characteristics of the distribution of the group before DIF analysis, general characteristics were determined by applying descriptive statistics to the whole group and the subgroups of other variables. As a result of the DIF analysis conducted according to gender variable, it was determined that different items showed DIF in MH and LR methods. The item found to show DIF by MH method was found to work in favor of girls. As a result of the DIF analysis conducted according to socioeconomic level variable, three items demonstrated DIF in MH method and five items demonstrated DIF in LR method. In both methods, it was determined that the two items showing DIF worked in favor of students with a low socioeconomic level. As a result of the DIF analysis conducted according to the residential area of the school variable, the school in the county and the school in the city were analyzed. The same two items demonstrated DIF in both MH and LR methods. In both methods, it was determined that one of these two items showing DIF works in favor of students who are in the city of the residential area where the school is located, while the other works in favor of students who are in the county. In the analysis of the school in the county and the metropolis, one item showed DIF in MH method and two items showed DIF in LR method. It was determined that the item that showed DIF in MH method also showed DIF in LR method and it worked in favor of the students in the metropolis where the school is located. In the analysis of the school in the city and the school in the metropolis, while six items showed DIF in LR method, no item did not show DIF in MH method. When MH and LR methods were examined in terms of the number of items showing DIF, it was seen that the two methods had different results and LR method determined a higher number of items showing DIF. In general, a positive and moderate correlation between DMF size rankings between the two methods were determined.