What is Latent Transition Analysis (LTA)?
Latent transition analysis is an extension of LCA in which you estimate the probabilities of transitions among behavior patterns over time. In an LTA, you estimate an LCA at each time point (hoping that the latent class structure is identical or at least highly similar at each time point) and additionally estimate the probability of transitioning from each class at one time point to all others at the next time point. In this way, LTA allows you to estimate movement among latent classes across shifting contexts, states, or developmental levels.
To use a smoking example, we could identify the number of patterns of smoking behavior in one week (based on a categorical indicator of smoking heaviness and another indicator of smoking volatility across days in that week), and then look at the same patterns of smoking heaviness and volatility in the next week. In an LTA, we would get estimates of latent class prevalences and smoking probabilities within each class for the first week and, separately, for the second week. We would also get an estimate of the probability of transitioning between classes in week one and week two. For example, we could estimate how likely it is that a week-one abstainer will transition to light intermittent smoking, light consistent smoking, or heavy smoking in week two. Covariates are also permissible in LTA and we could include treatment or other covariates to examine the degree of association between these variables and latent transition probabilities.