Reghdfe

[Serena Gallo]

Hi everyone,
I have a question.
My dependent variable is a score ranging from 0 to 100. It is a continuous variable. I want to include both country and industry-fixed effects.
Could I use reghdfe regression?

[Daniel Feenberg]

Well, of course you can use reghdfe, but it may not be appropriate. Predicted values may be outside the range of 0..100 and more subtle econometric problems will be present. Note that Stata can handle thousands of independent variables so you my be better off just creating factor variables for country and industry and using -tobit-. Don't assume your problem is too large for -tobit- until you have given it a trial. You could also stick with -reghdfe- and transform the dependent variable to +/- infinity with:

gen newdep=invnormal(.01*olddep)

or some other transformation, whichever works best.

[Serena Gallo]

Thanks a lot. Can I include in xttobit firms-fixed effects?

xttobit y x1 x2 x3 i.firms

[Daniel Feenberg]

Factor variables are not transparent to the author of a Stata command. If you write your own .ado command, there is some additional work invoved to accept factor variables. So you are right to wonder if -xttobit- will accept them as Stata does not document which commands accept factor variables. I just tried -xttobit- out with an i.var and it appeared fine with that. So you win.

However, do you want a random effects model? That isn't what you asked for in the OP.