I'm attempting to find a solution for a latent class mixture model with 4 nominal and 6 continuous indicators. My sample size is 530. I'm finding that regardless of the number of classes my fit indices continue to decrease but my LRT become insignificant for the 5 class model.
Any suggestions would be greatly appreciated. Thanks!
This is probably a sign that a simple latent class model is not available but may need adjustments such as some residual correlations among the items. Also, you may want to settle on a solution where BIC no longer has large drops when increasing the number of classes, despite not finding a minimum for BIC. Interpretability is important as well. And with many classes, some may simply be small variations on other classes.