I start with 9 independent variables in my linear regression. However, I find that the overall F test value is not having significant P value and some variables are highly correlated.

Thus I discarded five independent variables and get both significant P value for the overall F test and no multicolinearity.

However, I am a bit afraid whether I am having any problem like omitted variable bias although I really like the result I have.

If there are no other additional variables I can use more, can I use this result for my work or should there be something done for this to be fixed?

  • $\begingroup$ Have you checked out-of-sample performance? $\endgroup$
    – SmallChess
    Mar 29, 2017 at 4:19
  • $\begingroup$ My major concern is whether I am exposed to omitted variable bias after discarding many variables to have both significant F-test and no multiconlinearity problem. $\endgroup$
    – Eric
    Mar 29, 2017 at 7:57


Your Answer

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