Table 1 summarizes fit indices and entropy values resulting from the LPA of 2- to 9-class models using full set of SNPs. The best log likelihood was not replicated in the ten-class model indicating that the model is unstable, and the result from ten-class model was not presented. The optimal number of classes was not obvious based on information criteria and the BLRT. Both AIC and BIC continued to decrease throughout the 2- to 9-class models, and all BLRTs resulted in significant p-values. Given that the optimal number classes was not clear in terms of information criteria and BLRT, in the following sections we focus our interpretation on the 2- to 4-class models based on our expectations from twin findings. The classes were well separated in all models, as evidenced by values of entropy ranging between 0.888 and 0.920.
