Significant results without significant correlation

#1
Hi,

I have a weird case - I am finding significant effect (in a regression) for a variable where there is no significant correlation between them. I always thought that variables should be correlated to statistically affect each other.

I was just wondering what may be going on here. Any thoughts?

Regards
 
#4
This phenomen is caused by a suppression effect. A variable may not correlate with the DV but does explain variance of the DV by suppressing the "unimportant" part of another IV.

When you run an univariate regression with just this particular IV the result should be the same...

Marc