What Can RMLCA Tell Me?
An RMLCA can tell you how many latent classes of patterns of behavior or responses over time emerge in your data. There is no single indicator or simple decision rule for determining the optimal number of classes to retain, but a balanced consideration of several factors and model fit indices can help you determine which solution is best for your purposes.
An RMLCA can further help you estimate how prevalent, or common, the particular patterns of behavior or responses are.
Examination of the item response probabilities within classes in RMLCA can help you understand and label the latent classes that emerge. In a study of behavior change, one might well expect to see some stable patterns (changers and non-changers) and some common unstable patterns (initial change followed by relapse, initial non-response followed by later change, and unstable and intermediate behavior). Close examination and plotting of item response probabilities by class will help describe and differentiate the classes.
Covariate analyses can help you estimate manipulation or treatment effects on behavior or response patterns (latent classes) and help identify individual differences or characteristics associated with latent class membership. Multiple group analyses can also be conducted to see if the same classes emerge across groups (e.g., men and women).
Distal outcome analyses can tell you whether longer-term or distal outcomes vary among the latent classes in the model. It is not simple to estimate the degree to which latent class membership can predict later outcomes, given the probabilistic nature of classifying individuals in classes and some problems in the estimation of these relations. Despite these challenges, there are current ways to estimate these associations and new and improved approaches are in development (see the Methodology Center website at Penn State http://methodology.psu.edu/ for more on this)