Assesing of measurement invariance and DIF in high school entrance exams

The aim of this study is to examine the measurement invariance and changing item function of the verbal part Turkish subtest of secondary education transition exams in terms of gender and school type variables, respectively. The sample of the study is the random sample group selected by the Ministry of Education, General Directorate of Assessment and Examination Services, from the universe of this research, for students who choose to take the LGS held at the end of the 2017-2018 academic year. Multiple Group Confirmatory Factor Analysis to test the measurement invariance of the Turkish subtest; for DIF analysis, MantelHaenszel (MH), Logistic Regression (LR) and SIBTEST methods were used. Factor10, TAP, SPSS, R Studio and Mplus7 programs were used during the analyzes. Before the analyzes, the unidimensionality of the model created for the Turkish lesson using exploratory factor analysis was tested and confirmed by confirmatory factor analysis. The Turkish model, which was tested as a result of the measurement invariance analysis of the study, provided all stages of measurement invariance among student groups formed according to students' gender and school types. Because of the variable item function analysis, according to gender according to the MH method, items 1, 7, 12, 15 according to the MH method and in item number 15 according to the SIBTEST method. According to the type of school, DIF level B was observed in items 6, 8, 10, 16 according to the SIBTEST method. The findings of MH, LR, SIBTEST methods were found to be compatible with each other in the analyzes conducted according to gender. In the analyzes performed according to school type, it was found that the compatibility of MH and SIBTEST methods with each other was high, and the compatibility with the LR method was low.