R help - Jul 2002 - scale parameter and parameter vac-cov matrix in GEE

Dear all,

It looks like the parameters var-cov matrix returned by gee() is not
adjusted for the scale parameter:
> fm1 <- gee(nbtrp ~ strate * saison + offset(log(surf)), family =
poisson, data = Eff2001,
+             id = loc, tol = 1e-10, corstr = "exchangeable")
[1] "Beginning Cgee S-function, @(#) geeformula.q 4.13 98/01/27"
[1] "running glm to get initial regression estimate"
[1] -0.7755188 -1.5240927  0.4769241  0.4682660  0.1776812
[6]  0.1219127  0.6949677 -1.0451064
> fm2 <- gee(nbtrp ~ strate * saison + offset(log(surf)), family =
poisson, data = Eff2001,
+             id = loc, tol = 1e-10, corstr = "exchangeable",
scale.fix
= T, scale.value = 1)
[1] "Beginning Cgee S-function, @(#) geeformula.q 4.13 98/01/27"
[1] "running glm to get initial regression estimate"
[1] -0.7755188 -1.5240927  0.4769241  0.4682660  0.1776812
[6]  0.1219127  0.6949677 -1.0451064
> fm1$scale
[1] 10.01990
> fm2$scale
[1] 1
> 
> round(fm1$robust.variance, 4)[1:2, 1:2]
            (Intercept) strateLD
(Intercept)      0.0295  -0.0165
strateLD        -0.0165   0.0827
> round(fm2$robust.variance, 4)[1:2, 1:2]
            (Intercept) strateLD
(Intercept)      0.0295  -0.0165
strateLD        -0.0165   0.0827
>
What should I do to make the adjustement ?

Thanks in advance,

Renaud



-- 
Dr Renaud Lancelot, v?t?rinaire
CIRAD, D?partement Elevage et M?decine V?t?rinaire (CIRAD-Emvt)
Programme Productions Animales
http://www.cirad.fr/presentation/programmes/prod-ani.shtml

ISRA-LNERV                      tel    (221) 832 49 02
BP 2057 Dakar-Hann              fax    (221) 821 18 79 (CIRAD)
Senegal                         e-mail renaud.lancelot at cirad.fr
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

On Tue, 16 Jul 2002, Renaud Lancelot wrote:
> Dear all,
>
> It looks like the parameters var-cov matrix returned by gee() is not
> adjusted for the scale parameter:
<snip example>
>
> > fm1$scale
> [1] 10.01990
>
> > fm2$scale
> [1] 1
> >
> > round(fm1$robust.variance, 4)[1:2, 1:2]
>             (Intercept) strateLD
> (Intercept)      0.0295  -0.0165
> strateLD        -0.0165   0.0827
>
> > round(fm2$robust.variance, 4)[1:2, 1:2]
>             (Intercept) strateLD
> (Intercept)      0.0295  -0.0165
> strateLD        -0.0165   0.0827
> >
>
> What should I do to make the adjustement ?
You shouldn't.

Fixing the scale affects only the model-based variance estimate ("Naive
S.E." in summary.gee). This is done correctly.

 Setting up the robust variance estimate so that it relied on the fixed
scale would be a) difficult and b) pointless.  The whole reason for the
robust variance estimate is that it depends only on the mean model and on
the independence of different groups.


	-thomas




-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

I wanted to scale the var-cov matrix to compute confidence intervals for 
predicted values with a Monte Carlo procedure (simulation of predicted 
values assuming a multivariate-normal distribution for the GEE-model 
parameters). I incorrectly thought that I had to adjust the robust 
var-cov matrix with the scale parameter. Thanks for pointing out my 
mistake.

Best regards,

Renaud

----- Message Original -----
De: Thomas Lumley <tlumley at u.washington.edu>
Date: Mardi, Julliet 16, 2002 5:14 pm
Sujet: Re: [R] scale parameter and parameter vac-cov matrix in GEE
> On Tue, 16 Jul 2002, Renaud Lancelot wrote:
> 
> > Dear all,
> >
> > It looks like the parameters var-cov matrix returned by gee() is not
> > adjusted for the scale parameter:
> <snip example>
> >
> > > fm1$scale
> > [1] 10.01990
> >
> > > fm2$scale
> > [1] 1
> > >
> > > round(fm1$robust.variance, 4)[1:2, 1:2]
> >             (Intercept) strateLD
> > (Intercept)      0.0295  -0.0165
> > strateLD        -0.0165   0.0827
> >
> > > round(fm2$robust.variance, 4)[1:2, 1:2]
> >             (Intercept) strateLD
> > (Intercept)      0.0295  -0.0165
> > strateLD        -0.0165   0.0827
> > >
> >
> > What should I do to make the adjustement ?
> 
> You shouldn't.
> 
> Fixing the scale affects only the model-based variance estimate 
> ("NaiveS.E." in summary.gee). This is done correctly.
> 
> Setting up the robust variance estimate so that it relied on the 
> fixedscale would be a) difficult and b) pointless.  The whole 
> reason for the
> robust variance estimate is that it depends only on the mean model 
> and on
> the independence of different groups.
> 
> 
> 	-thomas
> 
> 
> 
> 
> 
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._