0
$\begingroup$

My correlation matrix is not showing very high correlations (less than 0.5), but still, when I am regressing panel fixed effects models, all my variables are omitted, and it shows it is due to collinearity among the variables. What is the way out of it?

$\endgroup$
  • $\begingroup$ This site must already have a number of closely related questions. Please search multicollinearity, singularity. $\endgroup$ – ttnphns Jan 17 '14 at 13:27
1
$\begingroup$

Let's say you have mulitple observations per person and you add person fixed effects, than you control for everything that is constant within a person like sex and race. As a consequence you cannot add these variables to your model, as you already control for them.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.