In contrast with most studies on detection of different item functioning (DIF) items, instead on the modeling and explanation of DIF. From the ideas of latent DIF and secondary dimensions, a mixed dimensionality model was proposed in this study on explanation of DIF. Using a non-DIF one-dimensional mixture 2-parameter logistic model as the baseline model. Six models were intentioned along with fully one-dimensional (Model 1a & 1b), mixed one-dimensional and two-dimensional (Model 2a & 2b), and fully two-dimensional (Model 3a& 3b) for two latent groups (non-DIF/DIF). The performance of the proposed model was evaluated by compared with different models by two real data analysis.
Comments & Questions:
1. Real data analysis is helpful for evaluating the performance of a new proposed model or method. But, whether a simulation study should be conducted prior to the data analysis?