This study presented a complex method to deal with the common educational investigation. Majority researchers were interested in the effect of teaching occasion in the student’s achievement. This transition analysis was modeled by author with a mixture IRT measurement model and multilevel latent class model for investigating the effect of an intervention. After introducing the model and the approaches and constraints of estimate, it was applied in an empirical data. The latent class model for illustrating the students’ ability in student level and the teaching style in teachers’ level was used to determine the categories by the model fit analysis. The process of this analysis and the practical interpretation reflected to the meaning of parameter were described.
1. In the traditional method, ANOVA and MANOVA were generally used to investigate the effect of intervention in experiment. I think the comparison between those traditional analysis and this complex transition analysis is necessary since we should want to know whether the result of new method is different and more sensible than the previous method.
2. In the paragraph of estimation, the label switching problem was mentioned. The problem can be solved by using constraint. But I didn’t understand how does this effect be occurred in MLE? Does it mean the pattern of latent class would be randomly matched to the class order in latent class analysis?