Latent transition analysis in Stata

[Liyuwork Dana]

Hello scholars,

I have seen that the question "Can Stata do latent transition analysis?" was asked by one of the Stata Forum member a couple of years before. But, because so many changes happened in Stata software over the last few years, I just wanted to make sure that this analysis can be done in the latest version of Stata. So can anyone tell me whether we can conduct latent transition analysis in Stata version 15.1 ( the recent version of Stata)? If yes, any special syntax we need to use because I haven't seen any procedure or syntax for this analysis.

Many thanks in advance,

Liyu

[Liyuwork Dana]

Hi again,

Continued from my previous question. Is there PROC LTA plugin function that can be used in Stata to conduct LTA?

Thanks,
Liyu

[Weiwen Ng]

Hi again,

Continued from my previous question. Is there PROC LTA plugin function that can be used in Stata to conduct LTA?

Thanks,
Liyu

I'll express interest in the answer also.

For those of us who don't know, latent transition analysis is basically latent class analysis with a time variable, and you can additionally estimate how the sample transitions between classes over time. This would be sort of like Markov modeling over latent, rather than observed, states. I would refer interested readers to this article by Lanza et al. PROC LTA was programmed by the Penn State University Methodology Center, and to my knowledge, they have not done an equivalent for Stata.

It seems like conducting LTA in -gsem- should be possible, drawing on -gsem-'s ability to fit latent growth curves with the data in wide format. However, that model treats the intercept and slope as continuous latent variables, and currently, Stata can't fit models that have both categorical and continuous latent variables (the latent class is, obviously, a categorical latent variable). It also isn't immediately obvious to me how to obtain the transition probabilities.

If fitting an LTA model is feasible in Stata 15, and you know the syntax, I would love to hear it. If it is not possible, I would respectfully request that StataCorp make this possible in Stata 16, complete with an example. If nobody can come up with syntax, then I shall add this to the Stata 16 requests thread.

[Leonardo Guizzetti]

I am not familiar with LTA/LCA but the UPenn Methodology Center does now appear to have an LCA plugin for Stata. The information and download are here.

[Weiwen Ng]

I am not familiar with LTA/LCA but the UPenn Methodology Center does now appear to have an LCA plugin for Stata. The information and download are here.

Correct, Penn State has a latent class analysis plugin for Stata, which is useful if you don't have version 15. To my knowledge, they do not have a latent transition plugin for any Stata version.

[Anna Manzoni]

I will also express interest in LTA in stata. Version 16 is now out but from my search I could not find anything about LTA. Any suggestions are appreciated if there is a way in Stata