Confidence intervals for what? For a Poisson glm() you can predict the
mean response or the linear predictor, and they are not the same (unlike a
linear model).
I suggest you predict the linear predictor (the default), use the s.e.s to
for a confidence interval and then if desired transform (exp in your case,
family(object)$linkinv() in general.). The code would be very similar to
predict.lm, and indeed if you study predict.glm you will see how to
call predict.lm to do this.
The family should not affect how you do it: poisson vs quasipoisson merely
affects the estimates of the s.e.s.
On Mon, 11 Aug 2008, Thomas Lapeyre wrote:
Hello,
I'm fitting a Poisson GLM with the glm( ) function and I would like to know how
to obtain the confidence intervals for predictions (fitted values)...
I mean like in function lm( ):
prediction.matrix=exp(predict(model1.lm,interval="prediction")
(where model1.lm is assumed to be a log-linear model fitted with the lm()
function.)
Is there any function which can do it? If not how can I compute the
prediction intervals from the fitted values? Is it the same for
"quasipoisson" models?
Thanks you very much!
Annexe: for example I use:
model2.glm=glm(Y~X1+X2, family="poisson")
_________________________________________________________________
[[elided Hotmail spam]]
[[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.
--
Brian D. Ripley, [EMAIL PROTECTED]
Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/
University of Oxford, Tel: +44 1865 272861 (self)
1 South Parks Road, +44 1865 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595
______________________________________________
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.