Can you direct me to an example, either in print or other, that uses MPlus to estimate the effect of one categorical latent variable on a second categorical latent variable? Hopefully this would include information about the nature of the latent variables (i.e., logit parameters of thresholds and/or conditional probabilities) and some type of regression parameter that links classes across variables.
It sounds like you are interested in latent transition analysis (LTA). See Examples 8.13 and 8.14 in the Mplus User's Guide. Also, see the section in Chapter 13 Parameterization of models with more than one categorical latent variable and the LTA references on the website.