Imagine 2 models: 1 for production, 1 for revenue, have cross-section and time dummies. Is it permissable to use CS in one and time in the other? How do I justify this? F-test for fixed effects is indicating this and also dummies I want to omit in each case are not really significant.
$\begingroup$
$\endgroup$
2
-
$\begingroup$ why would you want to do that ? $\endgroup$– user603Sep 29, 2010 at 22:36
-
2$\begingroup$ General remark -- try to give questions reasonable, informative titles and make them clear and understandable for people not working in your field. Also good question has a bigger chance for a satisfying answer. $\endgroup$– user88Sep 29, 2010 at 23:35
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
It seems that you are fitting simultaneous equation model in panel data setting. In this case using different dummies for different equations is entirely appropriate from statistical point of view. The justification should come from economic model.
The thing you should worry most is finding appropriate instrumental variables for your endogenous variables. Simultaneous equation estimates for panel data are covered in Baltagi book "Econometric analysis of panel data". You can also use Arrelano-Bond type estimates which are covered in the same book.
