Group Sequential Methods for Detecting Differential Item Functioning

Ensuring test fairness is a key goal of psychological measurements. This study focused on detecting Differential Item Functioning (DIF). DIF occurs when an item behaves differently for subgroups of examinees with the same level of latent traits. Item Response Theory (IRT) provides a framework for modeling these traits but often requires a large number of participants. This study addresses two research questions to resolve the conflict between theoretical requirements and practical costs. The first question investigates how to design sample sizes specifically for DIF detection within the Generalized Linear Mixed Model framework. The second question is whether Group Sequential Methods can reduce the required sample size. Group Sequential Methods are adaptive procedures that allow researchers to analyze data at interim stages. Through simulation studies, this study demonstrated that these methods effectively allow for the early termination of data collection when the DIF is present. The results showed that the proposed approach reduced the expected sample size while maintaining the Type I error rate at the nominal level. The discussion highlights that this methodology offers a cost-efficient strategy for test developers to identify biased items without collecting the full dataset.