You could run a fixed-effects regression (-xtreg, fe-) and interact other regressors with the gender-dummy. It's not a perfect solution, since you do not obtain a "standalone" gender coefficient and the interpretation gets a bit tricky, but at least you will have something to show for.
As an aside, even if you find the one model that may somehow yield a statistically significant gender coefficient and those results are then the only you present, it would still not be very good research practice. A non-finding is a result as well, although maybe not a very popular or publishable one...
As an aside, even if you find the one model that may somehow yield a statistically significant gender coefficient and those results are then the only you present, it would still not be very good research practice. A non-finding is a result as well, although maybe not a very popular or publishable one...
Comment