Page 1 of 1

Moderating effect - negative interaction and positive simple effects

Posted: Mon May 04, 2020 5:03 am
by olivia.lee
Dear PLS SEM colleagues,

I have a confusing result from my PLS analysis (in my view)

To make it simple,

- the model hypothesizes A influences B, A -> B. I have a moderator C.
- I have run a path, A -> B. It is positive and significant (my hypothesis is supported)
- An interaction term (e.g., A·C) on B is negative and significant (my hypothesis is supported)
- however, the simple effects C->B is POSITIVE and not significant (this is not hypothesized)


I wonder if this is disputable. I think the interaction effect and the simple effect should be negative but only the interaction terms is negative and the positive impact is a bit odd to me (or as the negative impact is not significant, it is useless to discuss?)

In fact, normally, C is regarded as having a negative impact on the dependant variable. This is the psychic distance. Larger psychic distance is associated with lower foreign production purchase intention (FPPI). So I expected that it will negatively affect the (FPPI) and the interaction (A·C) also negatively affect B. The independent variables are consumer characteristics such as social value shopping oriented, hedonic shopping motivation etc.

But the simple effect is positive (not significant). My question is that If I report this without any correction, it will invite any dispute? So confusing.

Actually, I have three independent variables but simplified as A. I modeled A1, A2, A3 affect B and C has moderating effects on A1-B, A2-B, A3-B.

I hope can have an advice from this forum.

Best regards,
OIivia