1. The model is quite complicated in modeling and used for simple division questions. The sample size of real data is extremely small. How come can the parameters be estimated well as illustrated?
2. In selection of latent class, it uses the multilevel 2-dimensional 2PLM to explore. It seems odd because it uses another model rather than our ‘interested’ model to do this. It needs more justifications.
3. The results obtained from mixture model may be less useful than cognitive diagnostic model in practical. We should focus on what the students cannot learn rather than how many classes are there.