You are not logged in. You can browse but not post. Login or Register by clicking 'Login or Register' at the top-right of this page. For more information on Statalist, see the FAQ.
It's a bit dangerous outside the basic model to simply include dummy variables to capture the fixed effects. Unless it's been show to be consistent that should not be assumed. Katie, I think, had no exclusion restrictions, and so an SEM is not appropriate.
My main gripe with reg3 is that it has no option for robust standard errors. One would like standard errors that are robust to system heteroskedasticity or serial correlation (like newey-west) or cluster correlation. reg3 is behind the times. Virtually every estimation method can be combined with robust standard errors. I don't see there's a way to do traditional 3SLS with the gmm command without basically programming the entire thing yourself. I would add that to a Stata 17 wish list.
Is there a problem for consistence in estimating by GMM 3SLS a simultaneous three-equation model for a panel data, for instance, with T=17 years and N=10 cross-section units? Is there a problem for estimation consistence here as a result of T being bigger than N?
Comment