Hello Guys,

I have a model in which variable m moderates 5 different independent variables (x1,x2,x3,x4, x5). When I examine only the direct effects y= x1+x2+x3+x4+x5+m (sorry for making the equation that simple), I get reasonable outcomes. However, when I start including the interaction term, let's say: y=i.x1##c.m+x2+x3+x4+x5+m, the previously significant and positive coefficient of a becomes negative and insignificant (which makes no sense logically). For including the single interaction terms, I used factor double factor notation. When I put the full model together, I used single factor notation: y=x1+i.x1#c.m+x2+c.x2#c.m+x3+i.x3#c.m+x4+i.x4#c.m+ x5+i.x5#c.m+m. In the full model, I also get weird and messy outcomes.

When I check the VIF, I get really high numbers such as 70+, until in the normal model everything is below 1. So think, I am not using the correct method to test moderation as the interaction term should not correlate with the normal term (causing a relly high VIF)


Can somebody please help me out?

Thank you so much!