Multicollinearity....

Zzzz

New Member
#1
... ok, anyone's thoughts on this would be great. I have a result in which my two IV predictors cancel out an IV, DV relationship. Thanks to John for giving me this new word!...but I'm confused at how this can happen. I understand that it can happen when the IV's are so strongly related, however it seems that mediational regression is perhaps not very robust? I wondered if there was a more sensitive test out there for mediation or is this the only one? i.e. I' m not satisfied that the M I've identified has no mediation value on the IV, DV.

I've heard of the Sobel test, I wondered what this is and how it differs from the 'normal' mediational equation, i.e. what exactly is it measuring?

So...theres two questions there: why is multicollinearity happening? - possible reasons?...and what does the Sobel test do thats different?

Thanks....
zzzzzzzz