In previous researches, the multiple indicator multiple cause (MIMIC) models are limited in used for detect the uniform differential item functioning (DIF). However, DIF has two types: uniform and non-uniform. When non-uniform DIF exists, then the MIMIC models can’t function. Thus, this paper added an interaction term in the MIMIC model, so-called MIMIC-interaction model, to test the non-uniform DIF. The interaction term is describing the relationship between item response and latent indicator depends on the level of latent variable ( θ ). A series of simulation studies were conducted to investigate the performance of the MIMIC-interaction model compared to the IRT-LR-DIF and the traditional MIMIC model. The results show that the MIMIC-interaction model can detect the uniform and the non-uniform DIF simultaneously, but cause the inflated type I error. The correction methods for the inflated type I error were suggested in the finally.
Future study:
1. Using the purification procedure into the MIMIC-interaction model and evaluating the performance.