Latent Class Approach to Detect Differential Item Functioning: PISA 2015 Science Sample

This study aimed to compare the performance of latent class differential item functioning (DIF) approach and IRT based DIF methods using manifest grouping. With this study, it was thought to draw attention to carry out latent class DIF studies in Turkey. The purpose of this study was to examine DIF in PISA 2015 science data set. Research Methods: Only dichotomous items were considered in this study. Turkey and Singapore samples were used to examine DIF. There were 6115 students in Singapore data set and 5895 students in Turkey sample. To detect DIF among countries based on manifest grouping, Item Response Theory Likelihood Ratio (IRT-LR) and Lord’s Chi-Square techniques were used. Besides, with Mixture Item Response Theory latent classes were defined and DIF items were detected with Mantel Haenszel method (MH) among latent classes. Number of DIF items were detected according to latent classes and the two countries were compared. Findings: There were 8 items including DIF among latent classes. With Lord’s Chi square method, four items were detected to include DIF at medium and high level among Turkey and Singapore. an‎d IRT-LR method revealed that only two items included DIF among countries. Implications forResearch and Practice: According to the results, it was recommended to use latent class approach in the investigation of DIF items in cross-country studies.