Soo Owen posted on Wednesday, October 03, 2018 - 5:45 pm
I am testing different LTA models with three time points data. LRT results indicate the model with first-order effect only is worse than the model assuming invariant transition probabilities (so, C1-> C2 transition probabilities = C2 -> C3 transition probabilities). Also, the model with both first-and second-order effect showed a better-fitting than the first-order model. Which model I would choose as the final model in this case? Or, do I also have to test time-invariance of the latent transition matrices for the second-order model in order to decide the final model?