A comparison of differential item functioning (DIF) detection for dichotomously scored items by using 2.1 IRTPRO, BILOG-MG 3, and IRTLRDIF v.2

This paper addresses statistical issues of differential item functioning (DIF). The first purpose of this study is to present an empirical data comparison of the IRTPRO, BILOG-MG 3, and IRTLRDIF programs and to detect DIF across two samples with IRT models, 1PL, 2PL, and 3PL. The second purpose is to examine IRTPRO to determine its effectiveness in detecting DIF, and, finally, to consider whether DIF exists in the GHSGPT for different ethnicities only in Social Studies. The GHSGPT predicts 11th grade students future performance on the Georgia High School Graduation Test and consists of 79 dichotomously scored items. The results show that several DIF items exist in the GHSGPT. For instance, all three programs consistently indicate that Item 13 is beneficial to Whites. In addition, IRTPRO is effective in detecting DIF because its results parallel those of IRTLRDIF and BILOG-MG 3.