Page 1 of 1
System of equation using panel data and gmm
Posted: Sat Jan 14, 2012 4:59 am
by tinfah
Dear Gareth,
How can I go about estimating a system of equations on panel data? Do I go into Object(system) first where I cannot see panel window and choose GMM; or
enter into estimate equation where I can see a panel window and choose GMM but I cannot estimate a system of equations in this.
Appreciate your reply. Thanks
System of equation using panel data and gmm
Posted: Sat Jan 14, 2012 8:21 am
by EViews Gareth
EViews doesn't support panel GMM for systems.
Re: System of equation using panel data and gmm
Posted: Sat Jan 14, 2012 9:08 am
by tinfah
Thanks for the quick reply. Any suggestion for software that can do this.
Re: System of equation using panel data and gmm
Posted: Fri Jan 20, 2012 7:44 am
by cel
Hi
can you tell me why in Eviews the 3sls system estimator doesn't return a J-statistic relating to the Instruments; the GMM system estimator does this.
Thanks
System of equation using panel data and gmm
Posted: Fri Jan 20, 2012 7:51 am
by EViews Gareth
No theoretical reason why it can't, it just hasn't been added to the output.
Re: System of equation using panel data and gmm
Posted: Fri Jan 20, 2012 8:00 am
by cel
Thanks for your reply. This highlightes that discriminating between different system estimators in eviews appears tricky:
FIML produces a loglik, but NLSUR doesn't.
GMM produces a J-stat, but 3sls doesn't.
Which statistical criterion/criteria would you then recommend to disriminate between estimating different nested systems a system - i.e., you estimate a system with 7 parameters and then a nested system with 5 parameters. Which one is better? The only thing i see is the Det Res Covar which i guess is part of the loglik calculation but not the full bit. I guess you could also do Wald parameter restriction estimates.