In its simplest form, PROC LCA allows the user to fit a latent class model by specifying a SAS data set, the number of latent classes, the items measuring the latent variable, and the number of response categories for each item. Multiple-groups LCA can be run using the GROUPS statement; users can examine measurement invariance across groups by adding the MEASUREMENT statement. Additional parameter restrictions can be provided in a SAS data file.
Continuous and categorical covariates can be included in the COVARIATES statement in order to examine the relation between each covariate and the probability of latent class membership. Prediction can be modeled using a baseline-category multinomial logit model or a binary logit model.
A Bayesian stabilizing prior can be invoked when sparseness is an issue for parameter estimation. Random starting values can be generated by the program, or the user can provide starting values in a SAS data file.
An empirical demonstration of PROC LCA appeared in Structural Equation Modeling:
Lanza, S. T., Collins, L. M., Lemmon, D. R., & Schafer, J. L. (2007). PROC LCA: A SAS procedure for latent class analysis. Structural Equation Modeling, 14(4), 671-694. PMCID: PMC2785099