KU STATLAB - Software



R package for latent variable models with categorical data


You can download an all-inclusive package for latent class analysis (LCA), latent transition analysis (LTA), and latent class-profile analysis (LCPA) here. This package fits a variety of latent variable models including:


  • LCA, LTA, and LCPA with multiple subgroups
  • LTA and LCPA with multiple latent classes (i.e. latent subgroups)
  • LCA with covariates to predict latent class membership
  • LTA with covariates to predict latent class membership at the initial time and transition between adjacent time points
  • LCPA with covariates to predict latent class-profile membership
  • LCA, LTA, and LCPA with flexible parameter restrictions


Estimation algorithms


  • Maximum-likelihood estimation using EM algorithm
    • Automatically use different sets of starting values to assure of global maximum
  • Bayesian estimation using Markov chain Monte Carlo (MCMC) algorithm
    • Optional dynamic data-dependent prior
    • Time-series plots for model parameters


You can learn about latent class analysis (LCA) and latent transition analysis (LTA) in this introduction provided by The Methodology Center at Penn State.


Download the package (written in R software)


  • Source files for maximum-likelihood estimation (CAT_LVM, Version 0.9.0 alpha): 
  • Source files for Bayesian estimation (CAT_LVM_BAYESIAN, Version 0.9.0 alpha):
  • Source files for data generator (GNR_DATA, Version 0.9.0 alpha):
  • Pre-loaded R Workspace (CAT_LVM, CAT_LVM_BAYESIAN, and GNR_DATA, Version 0.9.0 alpha):


User's guide and examples




Please send questions or comments about this package to here.