Date:         Wed, 20 Jun 2007 16:04:46 -0400
Reply-To:     Brandy Sinco <brsinco@UMICH.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Brandy Sinco <brsinco@UMICH.EDU>
Subject:      Re: Panel Data and Seemingly Unrelated Regression
Content-Type: text/plain; charset=us-ascii; format=flowed

Hi Lindsay,

If your panel data is approximately normally distributed,
your can use Proc Mixed.  If it is Poisson or binary, Proc GLINMIX.

HTH
Brandy
=====================================================================
From:    Lindsay Tuthill <lindsay.tuthill@ECONOMICS.OX.AC.UK>
Subject: Panel Data and Seemingly Unrelated Regression

Hello,

I'm trying to run a SUR model (a cost function and cost share equations)
with panel data, and I'm not sure which commands allow me to do that in
SAS.  I've run the pooled model using PROC MODEL, though I don't see where
PROC MODEL has a fixed or random effects option.  I've thought about
estimating the the SUR in deviations from the mean under PROC MODEL, but
this doesn't easily allow me to test for fixed versus random effects via
the Hausman test.  The only procedure that I'm aware of for panel data
specifically is PROC TSCSREG, but it doesn't seem obvious to me how to use
TSCSREG with a SUR model.  If you any suggestions/information as to how to
implement a SUR with panel data (fixed and/or random effects) in SAS, I'd
very much appreciate it.

Thanks very much,

Lindsay
====================================================================

Brandy R. Sinco, Research Associate II
U of M School of Social Work
www.umich.edu/~brsinco