EViews.com

Hi,

I am trying to run a linear regression with some time series variables. My explanatory variables are correlated probably because of which I am getting wrong signage on coffecients of certain variables. For example, for a regressor which is positively correlated with my dependant variable, I am getting a negative coefficient even though the regressor is significant. How can I control for multicollinearity without dropping any variable by checking VIFs?

Thanks,
Sreeja

The whole point of a running a multiple regression is to separate out the signs of the coefficients from the simple correlations between variables. Multicollinearity does not bias the coefficients.