In non-linear panel models, such as in a panel logit model, the incidental parameters problem may occur if T is small. From Cameron and Trivedi (2005), the incidental parameters problem may be addressed using a conditional maximum likelihood estimator, and is eliminated by conditioning that:

My question is that whether the fixed effects logit commands on stata such as "xtlogit, fe" or "clogit" already implements the conditional MLE process, and thus the incidental parameters problems is no longer a problem?


Absolutely, Stata's xtlogit, fe (or clogit, on which the former relies) estimates the maximum likelihood conditional on the sum of the outcomes $\sum_t y_{it}$, so that the incidental parameter problem is taken care of.

Please see here for a confirmation (slide 35) and some details. http://www.stata.com/meeting/wcsug07/cameronwcsug.pdf

  • $\begingroup$ Thank you! And I appreciate the slide deck source as well. This makes implementation on stata a lot easier. $\endgroup$ – Matt Lo Oct 6 '16 at 15:47

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.