Let's say I have a 3-way interaction and I perform a hierarchical regression, testing:
- Model 1 with only control variables.
- Model 2 including X and Y as well.
- Model 3 including the two moderators W and Z.
- Model 4 including the interaction effects (XW; XZ; WZ; and XWZ).
The relationship between X and Y is significant in model 2, but when I include the moderator variables (W and Z) it becomes non-significant, while the moderator Z and the interation effects XZ and XWZ are significant.
My question: Is there something wrong? What could be the theoretical and/or methodological explanation for that influence dissapearing?
The only idea that came to me is that the influence of the moderator variable Z is so strong that the direct effect of X on Y becomes non-significant. But I am afraid I might be doing something wrong and the model is not valid.
Thank you in advance!
Hello every one,
I was going to suggest that you refer to the literature for the choice of the selection of variables in the model in question to be sure of your methodology.
Second possibility is to apply the logarithmic function on the variable having an order of magnitude higher on the others knowing that the interpretation at the level of the coefficient obtained will change.
Third possibility is a methodology which already exists in statistics and which once I have used to apply in econometrics, it is to reduce and center all my variables to eliminate the difference in scale between them.
I admit having found significant variables by having confidence in this technique, very familiar in statistics.
Thanks and good luck !
- Badges
- Science topic
- Similar topics
- Statistics
- Survey