This page describes computer programs for latent class analysis and gives download links and/or contacts for requesting copies or obtaining more information.
Latent GOLDThis program by Jay Magidson and Jeroen Vermunt is simply splendid! With great graphics and intuitive commands, it includes new, state-of-the-art technical features like Bayes constants to help avoid boundary solutions, methods to diagnose and relax local dependence, and automatic testing of multiple start values. The features are too many to list here -- go visit their site for details.
Visit the Latent GOLD site and download a working demo at:
MplusAnother outstanding, cutting-edge program. Mplus, from Bengt and Linda Muthen, estimates a variety of mixture models (and other models), including LCA, latent profile analysis, mixtures of continuous variables, factor mixtures, and growth curve mixtures. Again, I won't even try to list all the capabilities of Mplus here. Definitely a program for serious researchers to consider.
For more information and to download a working demo, visit the:
LEMLEM is an excellent program by Jeroen Vermunt. It doesn't have as many features as Latent GOLD or Mplus, but it's free. LEM handles both unconstrained and constrained LCA (including ordinal variable, local dependence, and discrete latent trait models), as well as loglinear, latent trait, and other categorical data modeling methods. It checks model identifiability, and provides asymptotic standard errors for parameter estimates. Local dependence LCA models are easily handled.
Link to download LEM program, manual, and examples.
Department of Methodology
Faculty of Social Sciences
P.O. Box 90153 5000 LE Tilburg
MLLSAMLLSA (Maximum Likelihood Latent Structure Analysis), written by the late Clifford Clogg, is a bit old but still a good way to learn LCA.
Download an earlier PC version of MLLSA; mllsa.zip (44k bytes; zipped with PKZIP) contains executable code, sample input/output files, and abbreviated instructions for use. Fuller instructions are given in an Appendix to Allan McCutcheon's (1987) monograph from Sage Publications.
Alternatively, jump to the web site of Scott Eliason to download the most recent version of MLLSA, as well as other programs of his CDAS (Categorical Data Analysis System).
LLCALLCA, for Located Latent Class Analysis, estimates probit unidimensional latent class models, as described in Uebersax (1993). This is a discrete latent trait model, similar to the logistic unidimensional latent class (e.g., Lindsay, Clogg, and Grego, 1991), but based on a probit, rather than logistic assumptions.
PANMARKPanMark is another excellent program for LCA. It features automatic generation of multiple start values and bootstrap methods to statistically compare models with different numbers of latent classes. Parameter estimate standard errors are supplied. PanMark also handles longitudinal latent class (latent Markov) models.
Netherlands Central Bureau of Statistics
Depatment of Statistical Methods
P. O. Box 959
WINMIRA 2001This program estimates latent class models, Rasch models and Rasch mixture models (Rost, 1990; 1991; Rost & von Davier, 1992). In addition, it can estimate Rost's (1985, 1987) models for latent class analysis with ordered rating categories and supersedes the LACORD program used for that purpose. For information, see the WINMIRA Home Page or contact:
LCAPAn easy-to-use program for latent class analysis. This program is somewhat like MLLSA. A big difference, though, is that it can automatically generage multiple sets of starting values. For more information or to download the program, go to the LCAP Home Page or contact:
Department of Psychiatry
Washington University School of Medicine
St. Louis, MO United States
PROC LCA and PROC LTAPROC LCA and PROC LTA are new SAS procedures for latent class analysis and latent transition analysis (LTA) developed by The Methodology Center at Penn State. These straightforward procedures make it possible to pre-process data, fit a variety of latent class and latent transition models, and post-process the results without leaving the SAS environment. PROC LCA and PROC LTA are free of charge from The Methodology Center's website. Features include:
PRASCHPRASCH, by John Grego (web page), estimates the unidimensional logistic latent class model described by Lindsay, Clogg and Grego (1991). I'll try to post a link or the author's contact information here.
Department of Statistics
University of South Carolina
Columbia SC 29208 USA
MultimixThis program handles latent class analysis, latent profile analysis, mixture estimation, and combinations of continuous and categorical variables. Available at Dr. Murray Jorgensen's web page.
GLIMMIX 2.0This user-friendly program estimates mixture-of-regression models and mixture models. It allows many types of data, including brand choice, purchase frequency, "pick any N", paired comparisons, inter-purchase times, and conjoint data. For more information, including a free demo version, visit the
WinLTAWinLTA is a free-standing Windows application for conducting latent class analysis and latent transition analysis (LTA) developed and distributed by The Methodology Center at Penn State. LTA estimates mixtures of discrete-time Markov processes using multiple indicators. Although The Methodology Center no longer supports this program (see PROC LCA and PROC LTA for more recent software options), WinLTA and the user's guide are still available for download free of charge.
NEWTON and LATUnlike MLLSA and PanMark, which use Goodman's (1974) EM algorithm for parameter estimiation, NEWTON (LAT is an earlier version) reparameterizes the LCA model and uses Newton-Raphson estimation. It is not quite as easy to use as PanMark and MLLSA, as one needs to set a problem up in terms of design matrices; but arguably, the same feature adds flexibility to the program.
For information, contact the programs' author:
Department of Statistics
2006 Sheridan Road
Evanston, IL 60208-4070
LCABINProgram LCABIN (2.00) I don't know much about this program.
Go to Latent Structure Analysis site
Go to Rater Agreement site
Last updated: 10 May 2012 (WINMIRA edit)
(c) 2000-2007 John Uebersax PhD email