Re: st: xtgee and corr(ind) vs corr(exch)

["Sarah A. Mustillo" <smustillo@psych.mc.duke.edu>]

Garry -

My experience is that the coefficients DO vary in a GEE model, depending on 
the type of correlation structure specified.  Changing the correlation 
structure changes the parameterization of the model.  Quoting from Hardin 
and Hilbe (2002):

"Only one additional scalar parameter need be estimated if we believe that 
the observations within a panel follow no specific order and that they are 
equally correlated.  Alternatively, we may hypothesize a more complicated 
structure under the belief that the observations within a panel do follow a 
specific order.  Here, we may require a vector of additional 
parameters...or an entire matrix of parameters..."


So, both the coefficients and the standard errors can change if you change 
the correlation structure.  That said, your difference sounds rather large 
and I have no explanation for that.

Hope this helps a little!

Sarah

--On Saturday, November 02, 2002 1:42 AM +1000 garryaa@unimelb.edu.au wrote:


> Dear Statalist,
>
> Why is it that a coefficient can differ by up to 4 standard errors when
> an xtgee with corr(ind) is fitted compared to an xtgee with corr(exch) ?
>
> I have about 20 clinics and 40 patients per clinic, with the one covarite
> varying at the patient level, and the id variable is the clinic. The
> intraclass correlation coefficient is about 0.4. My understanding is that
> the expectation is that the coefficient should not differ when a
> different correlation structure is fitted. The coefficient goes toward
> the null when the exchangeable structure is fitted.
>
> Kind regards, Garry
>
> Vet. Science, Univ of Melbourne
> g.anderson@unimelb.edu.au
>
> --
> garryaa@unimelb.edu.au
>
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/


Sarah A. Mustillo, Ph.D
Center for Developmental Epidemiology
Department of Psychiatry and Behavioral Sciences
Duke University School of Medicine
Box 3454
Durham NC 27710

919 687-4686 x234
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/