Date:         Thu, 3 Apr 2003 07:58:43 -0800
Reply-To:     Iyue Sung <isung@EPIDEMIOLOGY.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Iyue Sung <isung@EPIDEMIOLOGY.COM>
Organization: http://groups.google.com/
Subject:      Marginal Effects for Generalized Linear Models
Content-Type: text/plain; charset=ISO-8859-1

Hello All,

  Does anyone know of any Macros floating about that calculates
marginal means for generalized linear models?

  For example, if I have covariates Age, Gender and Treatment (levels
A,B,C),
I'd like to estimate, say, the mean Proportion (or Incidence Rate) for
each Treatment level; analogous to a least-squares mean.

  I've searched quite a bit; but the closest I've gotten is that PROC
GENMOD, for example, will provide marginal log(Incidence Rates), with
the LSMEANS options.

  I'd like the marginal means on the original scale, and the
associated var./cov. estimates.  It wouldn't be too bad to write a
Macro I suppose (though I'm quite rusty with IML), or some code in S+
to do it 'by hand'.

But at the very least, seeing someone else's Macro, or if SAS already
has the capability that I'm not aware of, would be good confirmation
for what I believe the correct estimates are.


Many Thanks,
Iyue

-----------