Interaction effect vs. multi group analysis

Frequently asked questions about PLS path modeling.
Post Reply
PLS Junior User
Posts: 1
Joined: Tue Jun 07, 2016 1:23 pm
Real name and title: Timo Müller

Interaction effect vs. multi group analysis

Post by TimoM » Tue Jun 07, 2016 2:15 pm

Hello everyone,

if you would help me with my following questions I would really appreciate it.

I have a model with 2 dependent and 2 independent variables. Each of the 2 independent variables is linked to both dependent variables (so the model contains 4 paths). Now I would like to examine a possible moderation effect of a third independent variable on each of these 4 paths.

I read that in this case usually a multi group analysis is recommended.
But my problems are,that
(1) I have a very small sample size
(2) my moderator variable is continuous

I heard that both characteristics are arguments for examining a potentional moderator effect by adding interaction effects instead of running a multi group analysis.

My questions are:

(1) Is it reasonable in this case to examine the potential moderation effect by adding 4 interaction effects?

(2) If yes, do you have any useful literature verifying this?

(3) I read in this forum that it is recommended to check all interaction effects at the same time in one model and not each interaction effect in a separate model. Is this correct?

(4) If I decide to use interaction effects, do I have to create two models (one with the interaction effects and one without) like if I would use the MGA?

(5) If so, do I have to compare
- the model without the moderating variable with the model with the moderating variable (and its interaction effects)
- or the model with the moderating variable (but without the interaction effects) with the model with the moderating variable and the interaction effects?

Thank you very much, I really appreciate your help!

Post Reply