[R] Stats Question

[Ben Bolker]

David Winsemius <dwinsemius <at> comcast.net> writes:

> On Sep 13, 2010, at 8:27 PM, <Jennifer.Hollis <at> csiro.au> wrote:
> 
> > I have site data with variables that vary across sites (like wind  
> > speed, moisture content) and within sites I have experimental units  
> > (logs) with associated variables (like decay class, suspension).  
> > With normal response y one can use R to get the between and within  
> > site information using aov(y~u+v+Error(site)) where u and v are  
> > independent variables. Problem is we have a binary response (burn or  
> > not), so need to use glm. I need to know if it is possible to use  
> > glm with multistrata and if so how.

  If you want multistrata, you are looking at 'generalized linear mixed
models' -- for example in the lme4 package.

  glmer(y~u+v+(1|site),family=binomial)

  However, you should be aware that there are a lot of tricky issues
with GLMMs -- see the archives of the r-sig-mixed-models mailing list,
and <http://glmm.wikidot.com/faq> for a start.