Not a full answer, anyway:
 
To run Hausman's test you need to have a number of cross-units higher than a number of variables.
 
Good luck,
 
Bruno Thiago Tomio


De: gretl-users-bounces@lists.wfu.edu [mailto:gretl-users-bounces@lists.wfu.edu] Em nome de Julio Acuņa
Enviada em: segunda-feira, 7 de junho de 2010 14:12
Para: gretl-users@lists.wfu.edu
Assunto: [Gretl-users] * * Urgent question regarding Panel Regressions- Fixed and Random Effect

Sorry for the last email, dont know what went wrong.

Here´s my quesitons:

I´m running a panel regression, with 5 different units, 13 observations each for a total of nT=65.

1. When I test for fixed effects,  the test results shows no fxed effects presents, thus common intercept.  The thing is that with this regression everything is perfect: normally distributed errors, homescedastic erros, highly significant coefficiets, etc.  

With the above mentioned, according to theory I should be able to just run OLS ( pooled regrssion) because there are no fixed effects, But when I do that  I get heterescedastic errors,  i fix that  but I have to modify my model afterwards.


THe questions are:
1.  Is panel regression using some kind of robust estimation technique? such as newey west cov matrix?
2. WHen I test for random effects i get the following message: "not enough degrees of freedom", are we using the standard hausman test for random effects?

thanks for everyhtihng , salu2!!



Hotmail: Powerful Free email with security by Microsoft. Get it now.