I have conducted my interactive model using Partial Least Squares (PLS); software SmartPLS 2.0.M3.
On the one hand, regarding the interactions, how could we handle the dependence in error variance between the indicators of the main effect and the interaction construct?
Hi Manuel,
in LISREL we could include a correlation between error terms (Theta-delta), but in PLS we don´t have this possibility.
The only thing that I had thought was the assessment of the collinearity between predictor and moderator LVs.
Best regards.
Bido