I have a set of 6 DVs. They are change scores. I tried modeling them as the s and i of latent variables derived from the 6 at time 1 and the 6 at time 2, but the model fit was inadequate. I tried to model the difference scores with a latent change variable, and that didn't fit either. So it appears as if the 6 changes are distinct.
Yet, I don't want to test the effects of my ivs on each change score independent of the others. So I built a path model testing the effects of the iv and a mediator on all difference scores. The model output includes the intercorrelations among the DVs.
Here's the question: Does the effect of the direct and mediation path to each DV reflect prediction of the unique portion of each change score (controlling for the other change scores)?
And does the simultaneous test of all 6 variables in the same path analysis guard against alpha inflation by lowering degrees of freedom.?