Akademik Veri Yönetim Sistemi
International Journal of Assessment Tools in Education, cilt.13, sa.1, ss.95-107, 2026 (ESCI, TRDizin)
The COVID-19 pandemic has led to a widespread shift from traditional, supervised exams to unsupervised online testing environments, which has increased opportunities and motivation for cheating. Such dishonest behaviors threaten the validity and fairness of test results, underscoring the critical importance of robust test security measures. This study focuses on detecting individuals suspected of cheating by using two widely recognized statistical indexes: the 𝜔 and Generalized Binomial Test (GBT) indexes. Both indexes were applied within two analytical frameworks no-stage and two-stage methods. In the two-stage approach, the 𝜔 and GBT indexes were employed following the detection of potential cheaters using the Kullback-Leibler (KL) divergence index and person fit statistics (lz and lz*). To simulate realistic conditions, we manipulated key variables including test difficulty, ability levels of suspected copiers, and the proportion of copied items. Our findings demonstrate that the GBT index combined with the KL index consistently outperformed other methods across varying scenarios in terms of detection accuracy and control of false positives. These results suggest that integrating person-fit and distribution-based methods enhances the reliability of cheating detection in unsupervised testing environments. The study provides valuable insights for test administrators seeking effective statistical tools to safeguard test integrity, especially in the context of increasing online assessments.