Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: simple question about seemingly unrelated regression


From   Austin Nichols <[email protected]>
To   [email protected]
Subject   Re: st: simple question about seemingly unrelated regression
Date   Sat, 27 Jun 2009 17:51:06 -0400

The "different datasets" reply is irrelevant--see help suest which says:
"
Different estimators are allowed, for example, a regress model and a
probit model; the only requirement is that predict produce
equation-level scores with the score option after an estimation
command.  The models may be estimated on different samples...
"
What models are you running?  Country by country, or all countries in
one regression?  Does every model include the same explanatory
variables?  More info may result in better advice.  Also read the
manual entries on sureg and suest.

On Sat, Jun 27, 2009 at 5:37 PM, James Unnever<[email protected]> wrote:
> I just got a journal review back and a reviewer suggested that we use SUR.
> I have never used this procedure and know nothing about it.  However, I do
> not think it is appropriate.  I plan on sending the below to the editor.
> Could someone please tell me whether it is correct?
>
> Seemingly unrelated regression should be used when researchers run multiple
> models on the same set of observations thus generating the possibility that
> the residual errors from each equation may be correlated.  It would be
> appropriate to use this procedure (in SAS , Proc Syslin with the SUR option)
> if, for example, we were analyzing a dataset like the GSS that had multiple
> measures of punitive attitudes (support for the death penalty-should local
> courts be more harsh) and where running similar equations with each
> dependent variables (the equations cannot be identical).  In this case, the
> residual errors from each equation could be correlated and thus this
> procedure would be appropriate.
>
> This is not the case for our cross-national data.  Each of our regression
> runs are computed on a different data set -different observations.  Each
> dataset was independently collected in the countries we analyzed.  It may
> appear that the same observations are being analyzed because the same
> questions were asked, after the codebook was reinterpreted, in each country.
> However, each country has a unique identifier and an "if" statement must be
> included (e.g., if country = "England") to subset the data for that
> particular country.
>
> Thanks,
>
> Jim

*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index