Controlling for a variable in multiple regression creates significance of another var

#1
I have continuous variables A, B and C.
A/B correlates with C significantly. Neither A, B or A*B correlates significantly with C.

In multiple linear regression, with C as the dependent variable,
A is predicting C significantly, but only when B is controlled for.

A*B does not predict C in multiple linear regression.

What to make of this? Usually when you control for variables, significance of other variables drops. Here it is adding significance. I don't get it.
 

hlsmith

Omega Contributor
#2
Re: Controlling for a variable in multiple regression creates significance of another

Please don't duplicate posts. The issue is that some people will write in this thread while others post on the original - with both parties replicating content or not privy to the other thread.