Need help with Latent Transition Analysis I am new to this site, so my apologies if I am not asking the question the right way for this site. I am actually trying to understand Latent Class Analysis & Latent Transition Analysis. I have read a few articles. I found ProcLTA and ProcLCA in SAS, but in R I could only find poLCA package, and nothing for LTA. The webpage on which I found poLCA is below
http://sas-and-r.blogspot.com/2011/01/example-821-latent-class-analysis.html
Can someone please guide me how to do LTA in R and also what is the difference between Latent Class Analysis and Factor Analysis, as they seem very similar to me. Thanks in advance
 A: I've spent some time on this same question. I never found an R implementation, and I had a good deal of trouble uncovering the details of the SAS implementation. I think I finally found it in a technical note somewhere, but by that time I didn't need it anymore and I can't remember any of it now.
As far as the connection with factor analysis goes, it is factor analysis for categorical-by-categorical data. See here (page 5), among other places. Those slides also point out (although without any details) that LTA is also related to hidden Markov models. Page 19 of those slides list some free programs for fitting LTA models, including an R package that as far as I can tell doesn't exist and was never on CRAN. I did also just find an R package ltm that might be able to fit an LTA model. There's a JSS paper on it by Rizopolous (2006).
edit: In case you want to roll your own in R, technical details are in the PROC LTA User's Guide available here for free after you register an account.
A: I have been searching for a while for a LTA implementation in R, so this is what I found:
LTA is the name for a particular class of discrete-time Hidden Markov Models where both the latent and the manifest variables are categorical. So, as a matter of fact, you could use R-packages dealing with these class of models, in particular "depmixs4".
However, I recently found package "LMest" which is specifically designed for LTA. It includes some of the latest results on this specific field (see e.g. recent works of Vermunt or Bartolucci). 
For example:


*

*Both the manifest and latent distributions can be parameterized so
as to include individual covariates,

*Mixed LTA can be estimated
(i.e., a mixture of Markov Chains as mixing ditribution instead of a
simple Markov Chain),

*Viterbi's algorithm is implemented in order
to estimate the highest posterior latent statuses


The only con I have found is that it seems to me a little slower than proc lta in SAS
