Developments and Trends in Research of Detecting Multiple Group Differential İtem Functioning

Differential Item Functioning (DIF) analyses, which are critical for ensuring test fairness in education and psychology, necessitate the simultaneous comparison of more than two groups in contemporary measurement applications. This study aims to systematically elucidate the bibliometric and methodological characteristics of multi-group DIF detection methods by examining 37 foundational articles indexed in the Web of Science, Scopus, and ERIC databases. Employing a document analysis design, a qualitative research approach, the articles were subjected to content analysis in terms of research design, data type, group characteristics, statistical models, and software preferences. The findings indicate a distinct methodological evolution in the literature, shifting from classical frameworks such as the Generalized Mantel-Haenszel and Logistic Regression toward modern approaches, including Hierarchical Generalized Linear Models (HGLM), Bayesian estimation, regularization techniques (Lasso, TLP), and machine learning-supported algorithms. The analytical results demonstrate that while classical methods are often insufficient in controlling Type I error rates within complex and asymmetric data structures, innovative statistical models yield high detection power. In conclusion, it is determined that multi-group DIF research is increasingly trending toward analytical infrastructures that offer simultaneous and intersectional comparisons independent of a single reference group, and current methodological recommendations are provided for researchers.