Hi
This might be more of an econometrics question than a Stata question, although I'm looking for a way to do it in Stata (or someone to tell me that it can't be done).
I have a fixed effects model like the following:
yijt = Xitβ + αi + αj + εit + εjt
As you can see, there are two teams/individuals/firms (in my case, teams) in the equation, i and j. In my dataset I have them in two different variables, team1 and team2. This means that a team can be in team1 in one observation and in team2 in another. For absorbing fixed effects I just absorb each variable separately. I believe this is a loss of efficiency, but I don't believe it hurts convergence. (If someone has a smarter way to do this, I'm all ears as well.)
I expect there to be serial correlation in the error terms of each team, so I'd like to cluster on the team level. Is there a way to do this? Clustering on team1 and team2 ( -vce(cluster team1 team2)- ) allows for serial correlation for teams in each of the variables, but I expect my serial correlation to be across both the variables. Is there any way to handle this in Stata?
(It could just be that there's no way to handle this in econometrics in general. If so, please let me know.)
Thank you very much,
Rune
This might be more of an econometrics question than a Stata question, although I'm looking for a way to do it in Stata (or someone to tell me that it can't be done).
I have a fixed effects model like the following:
yijt = Xitβ + αi + αj + εit + εjt
As you can see, there are two teams/individuals/firms (in my case, teams) in the equation, i and j. In my dataset I have them in two different variables, team1 and team2. This means that a team can be in team1 in one observation and in team2 in another. For absorbing fixed effects I just absorb each variable separately. I believe this is a loss of efficiency, but I don't believe it hurts convergence. (If someone has a smarter way to do this, I'm all ears as well.)
I expect there to be serial correlation in the error terms of each team, so I'd like to cluster on the team level. Is there a way to do this? Clustering on team1 and team2 ( -vce(cluster team1 team2)- ) allows for serial correlation for teams in each of the variables, but I expect my serial correlation to be across both the variables. Is there any way to handle this in Stata?
(It could just be that there's no way to handle this in econometrics in general. If so, please let me know.)
Thank you very much,
Rune
Comment