Panel data regression fixed effects or random effects

[Frederik Tholstrup]

Hello everybody :-)

I'm currently working on a dyadic data set, which I want to analyze by using a panel data regression. I've already grouped my data, so my panel variable is called "pairid" (combination of country and sector) and "year" is declared to be my time variable.
xtset pairid year, yearly

So far, no problems. But when I use the command xtreg (...) , fe some of my independent variables are omitted and I guess thtat that's because they are dummy variables which happen to be constant for each pair, so Stata 'ignores' them. Then when I use the command xtreg (...), re cluster(country) instead the variables are not omitted and I get generally logic results, but they're not equal to the correct answer. I'm assuming that's because I use random effects instead of fixed effects. But again, as soon as I use fixed effects, many of the variables are omitted.

Does anybody happen to know a solution for this? Thank you in advance! :-)


Best regards,
Frederik

[Carlo Lazzaro]

Frederick:
what do you mean by "correct answer"?

[Frederik Tholstrup]

Hello again Carlo and thanks for answering,


my professor told me that the results I've gotten don't match with the results he has. And part of my project is to figure out the methodology as well, so I have to keep looking for all the right commands. I feel like I'm close, but I don't know if random effects really are the solution when dealing with panel data regressions. I wanna take care of the sector-fixed effects but with fixed effects some of the variables are omitted.


Best regards,
Frederik

[Sebastian Kripfganz]

That sounds like a question that only your professor can answer.

For some general background about constant variables in fixed-effects models, you can find a selected list of earlier Statalist discussions in the following topic: Fixed Effects and time-invariant variables.

[Frederik Tholstrup]

alright thanks! :-)