Evaluating WAIC and PSIS-LOO for bayesian diagnostic classification model selection