R help - Apr 2010 - Generalized Estimating Equation (GEE): Why is Link = Identity?

Hi,

I'm running GEE using geepack.

I set corstr = "ar1" as below:
> m.ar <- geeglm(L ~ O + A,
                 + data = firstgrouptxt, id = id,
                 + family = binomial, corstr = "ar1")

> summary(m.ar)
Call:
geeglm(formula = L ~ O + A, family = binomial,
    data = firstgrouptxt, id = id, corstr = "ar1")

 Coefficients:
            Estimate  Std.err    Wald Pr(>|W|)
(Intercept) -2.62516  0.21154 154.001   <2e-16 ***
ontask       0.00498  0.12143   0.002   0.9673
attachmentB  0.73216  0.35381   4.282   0.0385 *
attachmentC  0.25960  0.33579   0.598   0.4395
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Estimated Scale Parameters:
            Estimate Std.err
(Intercept)    1.277  0.3538

Correlation: Structure = ar1  Link = identity

Estimated Correlation Parameters:
      Estimate  Std.err
alpha    0.978 0.005725
Number of clusters:   49   Maximum cluster size: 533


Then, it shows that :
Correlation: Link = identity

Why is it not Link = logit?


Thank you,
Sachi

	[[alternative HTML version deleted]]

>From the GEE article in R News, Vol. 2/3, December 2002:
"Allows different covariates in separate models
for the mean, scale, and correlation via various
link functions."

Geepack offers link functions for the scale, correlation, and mean models.
 As the output suggests,

"Correlation: Structure = ar1  Link = identity"

does not refer to the mean link.  In fact, if you look at the output from
m.ar you would see:

"Scale Link:                   identity
Estimated Scale Parameters:  [1] 1

Correlation:  Structure = ar1    Link = identity"

See the R news article for more info on other correlation and scale link
functions.  The take home message is this: the mean link is exactly what you
think it is, the logit.

-tgs



On Thu, Apr 29, 2010 at 10:28 AM, Sachi Ito <s.ito.tcu@gmail.com> wrote:
> Hi,
>
> I'm running GEE using geepack.
>
> I set corstr = "ar1" as below:
>
> > m.ar <- geeglm(L ~ O + A,
>                 + data = firstgrouptxt, id = id,
>                 + family = binomial, corstr = "ar1")
>
>
> > summary(m.ar)
>
> Call:
> geeglm(formula = L ~ O + A, family = binomial,
>    data = firstgrouptxt, id = id, corstr = "ar1")
>
>  Coefficients:
>            Estimate  Std.err    Wald Pr(>|W|)
> (Intercept) -2.62516  0.21154 154.001   <2e-16 ***
> ontask       0.00498  0.12143   0.002   0.9673
> attachmentB  0.73216  0.35381   4.282   0.0385 *
> attachmentC  0.25960  0.33579   0.598   0.4395
> ---
> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
>
> Estimated Scale Parameters:
>            Estimate Std.err
> (Intercept)    1.277  0.3538
>
> Correlation: Structure = ar1  Link = identity
>
> Estimated Correlation Parameters:
>      Estimate  Std.err
> alpha    0.978 0.005725
> Number of clusters:   49   Maximum cluster size: 533
>
>
> Then, it shows that :
> Correlation: Link = identity
>
> Why is it not Link = logit?
>
>
> Thank you,
> Sachi
>
>        [[alternative HTML version deleted]]
>
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>
>
	[[alternative HTML version deleted]]