Moderating Variables

Frequently asked questions about PLS path modeling.
Post Reply
PLS Junior User
Posts: 8
Joined: Wed Aug 02, 2006 7:41 pm
Real name and title:

Moderating Variables

Post by tem3 »


I have a path loading from an exogenous LV (structure) to an endogenous LV (process) with a path coefficient of .73. When I add a moderating variable (Trust), I get the following path coefficients:

Trust main effect = .116
Trust x Structure interaction effect = .318
Structure main effect = .390

Why does my Structure main effect path coefficient drop from .73 to .39 when I add the interaction effect?

Thank you very much,
Teresa McCarthy
Teresa M. McCarthy, PhD
College of Business and Economics
Lehigh University
621 Taylor Street
Bethlehem, PA 18049
User avatar
PLS Super-Expert
Posts: 905
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL

Post by Diogenes »

Hi Teresa,

a possible explanation is the collinearity between the "interaction effect" and the others exogenous LVs.

In the first model all explanation is given by "Structure"
In the second model part of this explanation is divided with the collinear LV (interaction term).
(Did you center the variables to comput interaction term ? - SmartPLS have this option.)
(Are all indicators reflective?)

Just a idea:
1) Run the model with "Structure" and "Trust"
2) Run your 2nd model: "Structure", "Trust", and "Structure * Trust"
3) What happens with R2 ?

Best regards.

Post Reply