The objective of this study is to evaluate the effectiveness of the proposed GPCMlasso and Alignment methodologies in identifying Differential Item Functioning (DIF), by comparing their performances using real-world data. Specifically, the study focuses on the Interest in ICT subscale from the 2018 Program for International Student Assessment (PISA) ICT questionnaire. The analysis includes data from five Eastern European and Central Asian countries that participated in the PISA 2018 ICT questionnaire: Turkey, Bulgaria, Croatia, Georgia, and Kazakhstan, covering a sample of 29,277 15-year-old students. According to the Alignment method, the factor loadings for items 2, 3, 4, and 6, as well as the factor intercepts for items 2 and 6, remain consistent across all countries. In contrast, the GPCMlasso approach indicates that every item on the measurement tool exhibits both DIF and Differential Step Functioning (DSF) across various countries. The study finds that there is a 64% agreement rate between the two methods in detecting DIF. However, the GPCMlasso method appears to be more sensitive in identifying Differential Metric Functioning (DMF) compared to the Alignment method

Differential item functioning, alignment, GPCMlasso, machine learning, PISA 2018