I am working the first time with MPlus, so I am not totally sure about my model. I try to run a longitudinal mediation with 3 time waves, two mediators and two dependent variables. All variables are latent variables and I defined them before (not listed here).
T2M1 on T1M1 T1X; T2M2 on T1M2 T1X;
T3Y1 on T2Y1 T2M1 T2M2; T3Y2 on T3Y2 T2M1 T2M2;
T2M1 with T2M2;
Is the model so okay? All the variables are measures at all 3 waves, is it better to regress also the dependent variable T2 on the mediator T1, and also the mediator in T3 on the independent variable in T2 - or can I leave the model this way?
As default all the independent variables covariate. But in this model this leads to a covariance between the dependent variable in T2 with the independent variables in T1.
I guess that is not correct.
When I overwrite this command by using Model=NOCAVIRANCES command and specify both mediators in T2 as well as all the variables in T1 to covariate, the SMR raises a lot. Do I need any other ‘WITH’ commands?
So far my SMR increases a lot using the second way, but all the other fit indices are good. What does this mean?