I have a multiple regression model with moderate multicollinearity (VIFs <= 3.2). One of the coefficients is significant at the 10% level. Does this mean that despite of the variance inflation, I can still conclude that this coefficient is significant? I am asking because due to the fact that multicollinearity increases variance, and that the p-value is still significant, without multicollinearity it should be even more significant, no?

Or does multicollinearity not allow for any kind of interpretation like that?

Thanks a million!

  • $\begingroup$ Welcome to CV. Have a look here. You really should not be looking at $p$-values in the presence of colinearity. That said, there will always be weak colinearity when adding multiple variables. Maybe try assessing the severity with a better measure: the condition number. $\endgroup$ Jul 16, 2019 at 1:35
  • $\begingroup$ Related: stats.stackexchange.com/q/4099/176202 $\endgroup$ Jul 16, 2019 at 1:36


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.