How many anchor items should be selected in DIF-free-then-DIF strategy?

To control Type I error rates in assessing differential item functioning, constant item (CI) method was proposed in literature. The performance of CI method was found highly correlated to the number of anchor items, and several studies suggested four anchor items should be used when assessing DIF (Shih & Wang, 2009; Wang & Yeh, 2003). The strategy to select several items as anchor first and then assessing DIF for other items in the test was called DIF-free-then-DIF strategy (Wang, 2008). However, for short test, especially usually the case for the test with polytomous items, four anchor items might not be all DIF-free, and therefore the performance of DFTD strategy method could be fluctuated. In this study, “how many anchor items should be selected in DFTD strategy for polytomous items” was investigated within MIMIC method through simulation study. Four independent variables were included in this study: sample size, percentage of DIF items in the test, anchor item selection method, and number of anchor items. Our findings and conclusions will be given in details in the presentation.