Moderating variable

[Deepak Dass]

I am working on a model in PLS-SEM where I have a moderator variable that is hypothesized to moderate two different relationships between two mediating variables and the dependent variable (DV). Specifically, I have the following:

Mediator 1 (A) influences DV (D).
Mediator 2 (B) influences DV (D)
Moderator (C) is hypothesized (Theory) to moderate both of these relationships.

I want to know how to correctly model this in PLS-SEM, especially when dealing with interaction effects. Is it possible to use a SINGLE moderator for both relationships?
OR
Should I create two separate interaction terms, one for each mediator?
OR
Is there a better approach to modeling this scenario?

Additionally, I would appreciate any suggestions on whether to test these interactions separately or combine them into a single model. Would multicollinearity become a concern when testing two interaction terms simultaneously?

Thank you for any insights!

[jmbecker]

In general, it is recommended to model all of this in one model. Otherwise, you may suffer omitted variable bias.
However, that usually also requires slightly larger sample sizes because the complexity of the model is larger and there may be some more multicollinearity. Still it would be the best way to do it.