[R-sig-ME] glmer in R
David Duffy
davidD at qimr.edu.au
Thu Mar 10 09:55:46 CET 2011
On Tue, 8 Mar 2011, Leena Hamberg wrote:
> Thus, I would like to verify that I have understand instructions concerning
> glmer in R correctly. Is it so that you can - instead of quasi distributions
> - put your observation unit as a random factor into models?:
>
> model1=glmer(height~treatment+basaldm+treevolume+saplnum+browsing+(1|sampleplot)+(1|obs),family=poisson(link
> = "log"),niter=100,data=sorbaucu)
>
> Sample plot in the model above is a real random factor (including several
> observations) but variable "obs" means observation units (as a factor). Thus,
> does this handle the same thing as does quasi error distribution?
Yes, but better ;)
Actually it is not quite so simple, since the different models have
different properties that might be better or worse for mimicking the
particular biological or physical process. The paper by Ver Hoef and
Boveng (Ecology 2007) is one example (the negative binomial can be derived
as a poisson-gamma mixed model, while the glmer approach described is a
poisson-normal).
How does height get to be poisson, btw?
Cheers, David Duffy.
--
| David Duffy (MBBS PhD) ,-_|\
| email: davidD at qimr.edu.au ph: INT+61+7+3362-0217 fax: -0101 / *
| Epidemiology Unit, Queensland Institute of Medical Research \_,-._/
| 300 Herston Rd, Brisbane, Queensland 4029, Australia GPG 4D0B994A v
More information about the R-sig-mixed-models
mailing list