The optimum latent class model (by BIC model fitting criteria) was a 3-class model. The sample-size-adjusted (ssadj) BIC decreased from 13,849 for a 3-class model to 13,827 for a 4-class model, with a corresponding drop in entropy of 0.77 to 0.68. The ssadj BIC from the factor mixed model was 13,830, with an entropy of 0.768. Due to the similar results between the two models, a more parsimonious 3-class model which assumed conditional independence was used in subsequent analyses.
