Message-ID: <loom.20050805T175241-499@post.gmane.org>
Date: 2005-08-05T15:56:43Z
From: Dieter Menne
Subject: prediction from glm
Hugues Santin-Janin <hsantin-janin <at> gct.org.uk> writes:
> I try to fit birds counts over years using glm. I have done (with Estate
> and year as factors):
>
> Model1 <- glm(Females~Estate+Year+offset = log(area)), family =
> quasipoisson(link = log), na.action = "na.exclude")
..
> Pred1 <- predict(Model1, type = "response", na.action = "na.exclude")
>
> My question is: How can I obtain predictions for Females in each year
> that are standardized by averaging over the levels of Estate?
You should use the newdata argument in predict.glm to construct the "should-be"
data set. The example on the page ?predict.glm creates budworm-data on the fly,
but it's probably easier to understand if you create a separate data frame
myfemales first, the do predict(Model1, newdata=myfemales,..) in a separate
step.
Dieter Menne