Latent Class Goodness-of-Fit Statistics Guide

This guide details the process of obtaining goodness-of-fit statistics for latent class models, specifically focusing on the evaluation of model fit through likelihood-ratio tests. The document outlines the steps taken to fit a two-class model and compares it with a saturated model using the estat lcgof command. It provides specific fit statistics, including the likelihood ratio, Akaike’s information criterion (AIC), and Bayesian information criterion (BIC), and discusses the implications of these statistics in determining model fit. The guide also describes the process of comparing models with varying numbers of classes, referencing the work of Goodman (2002) for methodology. It includes results from fitting one-class and three-class models, along with their respective fit statistics and likelihood-ratio tests. The lcstats command is introduced for a more direct comparison of the models, reporting class separation and additional information criteria. Overall, the guide serves as a comprehensive resource for understanding and applying latent class analysis techniques.