1. I have never seen a research discovering latent classes as many as this study. Moreover, only limited descriptions were for uncovered latent classes.
2. Since a model with relatively better fit was desired, a series of comparison should end at the 9-class 8-dimension model, which had BIC = 68,653.32. Thus later analyses were redundant.
3. Was it suitable to reduce dimensionalities by using a non-IRT approach (hierarchical cluster analysis)?
4. Was it assumed in Equation 6 and Table 6 that a common covariance structure is constraint for all latent classes? If yes, such restriction may be rather unreasonable.