The studies develped a new method by adding a scale purification procedure to the rank-based method (RB-S) in which the DFTD strategy was used. Two simulations were conducted to compare RB-S with RB (simulation 1) and with the traditional DIF method (AOI, AOI-S, PA) where no DFTD strategy was used. It was found that RB-S method yield a well-controlled Type I error and a high power rate of DIF detection, especially when the ASA is far away from zeor (ie, high percentage of constant DIF items).
I summarized some main DIF assessment method here:
RB-S method: use RB-S method to select anchor and then assess the other items for DIF with the designated anchors (purification procedure, DFTD).
RB method: use RB method to select anchor and then assess the other items for DIF with the designated anchors (no purification procedure, DFTD).
AOI method: all other items were treated as anchors when assessing the studied item for DIF (no purification, no DFTD)
AOI-S method: the AOI method with scale purification procedure (purification procedure, no DFTD)
PA: pure anchor method, in which four clean anchors were selected by design and then the other items were assessed for DIF.
Limitation of IRTLRDIF in implement CI method: when the CI method is implemented with IRTLRDIF, the test is treated as if it consisted of the anchored items and the studied item. For example, when assessing item 1 for DIF with the four anchor, IRTLRDIF treated the test as if it consisted of the four anchors and item. And, if there is only one item was used as anchor, the test was treated as consisting of only two items. As a consequence, the result is very poor. Hence IRTLRDIF will reduce the power rate of the CI method, and thus the DFTD strategy.