Re: st: Testing cross-model hypothesis in long panel regressions.

[Svendsgaard.David@epamail.epa.gov]

You could do a meta analysis:

meta Effect  SE

where Effect  is a variable with length two and contains coefficient of
x11 and coefficient of x12


and where   SE   is a variable with length two and contains the SE of
coefficient of x11 and the SE of coefficient of x12.

But there is probably a better way.



David J Svendsgaard, PhD
Biostatistician
EPA/ORD/NCEA/RTP, Mail Drop B-243-01
Research Triangle Park, NC 27711
Phone (919) 541-4186
Fax (919) 541-1818


                                                                                                        
  From:       Jorge Eduardo Pérez Pérez <perez.jorge@ur.edu.co>                                         
                                                                                                        
  To:         <statalist@hsphsun2.harvard.edu>                                                          
                                                                                                        
  Date:       08/19/2010 02:41 PM                                                                       
                                                                                                        
  Subject:    st: Testing cross-model hypothesis in long panel regressions.                             
                                                                                                        
  Sent by:    owner-statalist@hsphsun2.harvard.edu                                                      
                                                                                                        





Dear Statalist,

I have two models estimated with panel corrected standard errors

xtpcse y1 x11 x12 x13, corr(psar1)
xtpcse y2 x21 x22 x23, corr(psar1)

And I want to test cross-equation hypotheses, for example, test if the
coefficients of x11 and x21 are the same. I've tried using -suest-,
but it is not available for panel corrected standard errors. I also
tried estimating the models using -xtgls ..., corr(psar1) nmk-  and
then trying -suest-, but this won't work because -suest- needs scores
in order to calculate the simultaneous covariance matrix. Also, I'm
not sure if the underlying assumptions for -suest- (i.e, the
coefficients being jointly assymptotically normal) are satisfied in
this case.

My current solution is to take the coefficients of the first of the
models as given and test if the coefficients of the second model are
equal to the coefficients of the first model.

Does anyone know how can I test cross-equation hypotheses in this
case? I apologize if this is more of an statistical question rather
than an Stata question.

Thank you.
_______________________
Jorge Eduardo Pérez Pérez

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/