Second, we confirmed the invariance of item probabilities ([rho]) over three time points and across groups using IC instead of LRDT due to the large degree of freedom, which helped to interpret the latent transition models, stabilize estimation, and improve model identification.
In model comparison through LRDT, we found that Model 4 (constraining the prevalence and transition parameters) was best fitted to the data because it was equally well fitted as compared with other models (p > .05) and was the most parsimonious.