I have a complex SEM model where series of latent variables are predicting zinb outcomes at two time points. Is it appropriate to use the zinb outcomes at t1 as mediators for the zinb outcomes at t2? If so, how should/could I go about conducting this? I am using Mplus.