Multiple Regression, Collinearity Question

Hello All,

I'm running a multiple regression on 8 predictors. A few of the predictors have positive correlations with the response, however their regression coefficients are negative. This seems off to me. I've done collinearity diagnostics (VIF, condition index) and there are no redflags. When calculating the condition index I do not use the intercept, otherwise, if I do a few of the condition indexes do explode, but when I don't use the intercept no condition index is higher than 3.

What can explain this positive correlation, negative regression coefficient situation? Thanks.


Ninja say what!?!
Its possible that adding in the other covariates is what's causing the coefficient to be negative. Try running independent regressions using only each (one) variable of interest. Let me know what happens.