This study mainly explain the latent classes of examinees' characteristic and DIF items in the multidimensional structure. The Author utilized the second dimension latent trait to indicate the latent DIF classes, such as the other respondent strategy. Then, The model 2 in Figure 1 is the proposing model in this paper. comparing with the model 1 explain the DIF in latent class; comparison between the 2a and 2b explain the DIF effect of the discrimination on the primary dimension; the comparison between model 2 and 3 explain whether the secondary dimension limited to the DIF latent class or the application on non DIF class is allowed.
1) I am curious about whether the definition of DIF conflicts with the conception of this multidimensional model. The Different Item Function should be illustrate as the same item has the effect of different parameter under examinee's variable. However, this paper use the second latent trait to indicate the DIF items, It seems a kind of fundamental changing even if the model fit is good.
2) I am not sure how did he discriminate the subjects belongs the DIF latent class or the non DIF latent class. In model 1 and model 2 we need use different formula for different latent classes, and the model fit method for comparison those model. did it detect the subject's respondent strategy in its applications?
3) I still feel strange that speedness test is one kind of DIF. The speedness effect, the subject speeded to response the last items in a test, is most like the guessing effect. Maybe I should read the literature listed by author for checking this effect.