Investigating Measurement Invariance of Ankara University Foreign Students Selection Test According to Latent Class and Rasch Model

Abstract

This study aimed to investigate the factor structure of the Basic Learning Skills Test included in the Ankara University Examination for Foreign Students and determine whether this test has measurement invariance and whether the items of the test shows differential item functioning according to the country and gender. Of the descriptive research models, screening was used in the design of the study. The population of the study consisted of 2134 individuals. Since the research involved intercultural comparisons, culture was taken into consideration in the selection of the sample. Accordingly, 1110 individuals from Germany, Azerbaijan and Iran constituted the sample. In the research, the first 60 items of the Basic Learning Skills Test concerning letters, numbers and figures were utilized. Latent Class and Latent Class Factor Analyses were employed to determine the factor structure of the test, and a Simultaneous Latent Class Analysis was undertaken to investigate the measurement invariance of the results according to country and gender. Besides, Recursive Partitioning Analysis based on Rasch Model was used to identify whether the items included in test show differential item functioning by country and gender. Before determining the factor structure of the Basic Learning Skills Test, an examination was conducted to ascertain whether the test as a whole fitted the scale structure, and which classes were included in the response patterns of the individuals taking the test. According to the results, the test had a latent model of four classes and a multiple-factor structure, comprising three factors of establishing part-to-whole relations, making inferences, and analytical thinking. Concerning the findings of analysis regarding the measurement invariance according to country and gender, it was found that the test had a homogeneous structure, and as a result, the structure measured by the test was equivalent across the groups compared. Studying whether the items in the test show differential item functioning according to country and gender, it was discovered that 14 items involve differential item functioning by country. Differential item functioning in regard to gender was not observed.