Hi everyone,

I'm conducting a Latent Transition Analysis (LTA) in Mplus across three time points, with continuous latent profile indicators and 4 continuous covariates. My goal is to assess how these covariates predict profile membership and transitions over time.

To avoid listwise deletion due to missing data on covariates, I followed Mplus documentation suggesting to include covariates as dependent variables—achieved by mentioning their variances in the MODEL command. However, Mplus then requires numerical integration, and only allows the MONTECARLO integration algorithm in this case. When I do use MONTECARLO, the estimation takes extremely long (over 24 hours) and still doesn’t produce any output.

I'm considering simplifying the model by running separate LTAs for T1→T2 and T2→T3 with one covariate at a time, but that would lead to different class solutions and varying sample sizes.

My questions:

Any advice, suggestions, or shared experience would be highly appreciated.

Thank you! Mahdi