forum.smartpls.com

Hi,

I have two predictors (p1,p2) and one moderator (m) that is moderating the relationships of p1 and p2 with my endogenous LV (E). To evaluate the moderating effects, can I make use of one model that contains both interaction effects (p1, p2, m, m*p1 and m*p2) or do I have to measure both moderating effects separately (p1, p2, m, m*p1 in a first, and p1, p2,m, and m*p2 in a second interaction model)?

Intuitively, I would measure both moderating effects in the same model since both effects exist in parallel. (Thus far, I have not identified any source that contradicts this approach)

(The question can be extended on another case: if I had two moderators that each influence the relation of one exogenous LV (m1 affects p1->E and m2 affects p2->E), would I measure both interaction effects in the same model or would I have to measure them separately?)

Thanks a lot,

Patrick

Hi Patrick,
in this issue, PLS has more in common with regression than LISREL.

For this reason, I think that you should work like stepwise regression.
1st) Just predictors
2nd) Including one moderator
3th) Changing the moderator
4th) Including all
With these results you could decide if there is collinearity too.
Best regards.
Bido