Re: [R] predict from glmer

[Ben Bolker]

sj  gmail.com> writes:

> predPN <- model.matrix(terms(mod.PN),newDat) %*% fixef(mod.PN)
> 
> this seems to work fine and the I print the predictions
> 
> print(PredPN)
> [1] [2]
> 2.358722 2.312340
> 
> However I am not certain what I am looking at here,  my best guess is that
> each of these could the odds (p/1-p). But I am not at all sure in my guess,
> any insights would be appreciated

   These are most likely to be log-odds rather than odds.
(gmane wants me to say more or quote less ... you can use
plogis(PredPN) to get the probabilities.)

__
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.

[R] predict from glmer

[sj]

Hi all,

Windows XP
R 2.10.0

I am trying to generate predictions from a binomial mixed model using glmer.
the models is as follows:

mod.PN <- glmer(PN_out~ as.factor(intr3) + transcode + Yr + (Yr|ID)
family=binomial,data=partset, na.action=na.omit)

where PN_out is a two column binomial variable

I then used the code suggested by BErt Gunter to predict from lmer:
here I am interested in making from two new predicting two new observations
where I hold everything fixed other than the factor intr3

predPN <- model.matrix(terms(mod.PN),newDat) %*% fixef(mod.PN)

this seems to work fine and the I print the predictions

print(PredPN)
[1] [2]
2.358722 2.312340


However I am not certain what I am looking at here,  my best guess is that
each of these could the odds (p/1-p). But I am not at all sure in my guess,
any insights would be appreciated


Best,

Spencer

[[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.