It comes from linearizing log(x) around the fitted value, derivative of log is
1/x, so
aa$se.fit/aa$fit == bb$se.fit
For diverged estimates, this gets dubious as the linearization only works in a
neighborhood and for x not too close to the singularity of log() at zero.
-pd
> On 02 Mar 2017, at 09:22 , Patrick Connolly <p_conno...@slingshot.co.nz>
> wrote:
>
>
> I'm trying to calculate a CI for predictions from a Poisson GLM object
> egg.glm.
>
> Browse[2]> aa <- as.data.frame(predict(egg.glm, newdat, type = "response",
> se.fit = TRUE)[-3])
> Browse[2]> bb <- as.data.frame(predict(egg.glm, newdat, se.fit = TRUE)[-3])
> Browse[2]> aa
> fit se.fit
> 1 6.144212e-07 0.0005114257
> 2 2.452632e+01 5.4657657443
> 3 1.440000e+01 2.5817126393
> 4 4.389796e+01 4.5533997800
> 5 3.820455e+01 4.4827326393
> 6 6.226667e+01 5.6589154967
> Browse[2]> bb
> fit se.fit
> 1 -14.302585 832.36979026
> 2 3.199747 0.22285311
> 3 2.667228 0.17928560
> 4 3.781868 0.10372691
> 5 3.642954 0.11733506
> 6 4.131426 0.09088194
> Browse[2]>
>
> bb$fit is clearly log of aa$fit but just what is se.fit? How do I use
> it to get a CI which is calculated on the log scale? The first one is
> a bit messy since it is entirely from zeros. Should I remove those or
> would that be unnecessary?
>
> TIA
>
> --
> ~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.
> ___ Patrick Connolly
> {~._.~} Great minds discuss ideas
> _( Y )_ Average minds discuss events
> (:_~*~_:) Small minds discuss people
> (_)-(_) ..... Eleanor Roosevelt
>
> ~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.
>
> ______________________________________________
> R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.
--
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Office: A 4.23
Email: pd....@cbs.dk Priv: pda...@gmail.com
______________________________________________
R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see
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.
