adjustment procedures in multiple testing was ordinarily used in hypothesis testing for control the Type I error rate. The effect of the 3 adjustment procedures (Bonferroni, Holms, Benjamini Hochberg) were investigated in the 4 DIF test method (Mantel Haenszel, Logistic Regression, Differential Functioning item and test, and Lord's chi square test) in this study. The Type I error rate and Power were the indices for evaluating those effects. The result showed the IRT based method (DFIT and Lord's) are not necessary to use adjustment procedures.
1. this study gave us the understanding about the phenomenon of the adjustment procedures used in DIF methods. However, there were no explaining for those results. For example, why the IRT base were not influenced by those procedure. Why those Type I error rate are underestimate in Bonferroni and overestimate in BH for MH and LR methods?
2. whether those adjustment procedures can be implemented in Hui-Fong's interaction DIF study?
3. In my intuitive before I read the detail, I thought the effect of adjustment procedures should be useful in the multi focal groups such as the race variable (white man, blacks, and Asian) in DIF test.