This article proposed a purification procedure on Rank-based method and compare it with RB method (Simulation 1) and other four methods which implemented by IRTLRDIF (Simulation 2). The results show that when the average signed area (ASA) is nearly zero, all methods’ performance are good. While the ASA apart from zero, the RB-S method yields the best performance among all methods investigated which well control Type I error rate and Power. The authors also pointed out the future research related issues: RB-S for polytomous items, procedures improvement in exiting DIF software, and nonuniform DIF situations, etc.
This paper presents an outstanding review of methods on assessing DIF, including the equal-mean-difficulty (EMD), the all-other-item (AOI) and the constant-item (CI) methods, as well as the DIF-free-then-DIF strategy and purification procedures.
The author also argued that the percentage of DIF items is not the case of inflated Type I error and deflated power of the above methods but the average signed area. But the author did not mention that how to evaluation ASA in practices. Since if the ASA equals 0, all methods are performing very well and the practitioners would have no idea to choose a method.
The EMD method have not been study might due to the IRTLRDIF software which do not implementing it. But it can be expected that it will have good performance when ASA equals 0.