Hi, In running a latent class analysis. Do I first check for significant categorical predictors and only use those in the latent class model? For example, if I have a dependent variable that is binary and I have 30 categorical independent variables, should I run a logistic regression first to find those variables that are significant predictors of the binary dependent variable. Then use all those plus the significant independent variables in the latent class analysis? -- Thanks, Jim. [[alternative HTML version deleted]]