we have estimated from two different sets of variables in one dataset two CFA models with two separate Mplus programs. Now my colleagues want to do regression with the factor scores obtained from both models. Should I have used one program to compute all factors from both models rather than doing it separately? In the latter case my impression is that the correlation between items belonging to different models can not be taken in account by Mplus. On the other hand it may play only a role for model evaluation?