> ----------
> De : [email protected]
> Envoy� : jeudi 9 janvier 2003 18:13
> � : [email protected]
> Objet : RE: GEE
>
> Xavier:
>
> Why would you want an autoregressive time-series correlation structure for
> this data? It appears to me that the exchangeable would be the one
> desired. The latter is the one that you would nornally want for
> longitudinal data, unless it is over a great many time frames.
>
> Joe Hilbe
>
>
>
Dear Joe,
Thank you for your rapid answer. I wanted to use autoregressive time-series
correlation structure after reading Davis (Statistical methods for the
analysis of repeated
measurements, Springer, 2002): "Although assumption of constant correlation
between any 2 repeated measurements may not be justified in a longitudinal
study, it is often reasonable in situations in which the repeated
measurements are not obtained over time. The AR-1 model is a natural one to
consider when measurements are taken repeatedly over time".
Finally, I will use the exchangeable correlation structure for experiment.
However I would be pleased to have my initial question answered: is it
possible using xtgee to take simultaneously into account two factors (for
example time and patients as in my data) ?
Sincerely yours,
Xavier Viviand
*
* 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/
