Charles Willis <willis.charlie <at> gmail.com> writes:
>
> Hello,
>
> I am running the program COMPAR.GEE within the package APE. My dependent
> variable is binomial, while my independent variable is a multi-state
> categorical variable. The output reports an estimate for each state of the
> independent variable except the first one. For example, for the variable X
> with 3 states, the output is:
>
> intercept (estimate)
> X2 (estimate)
> X3 (estimate)
>
> I have two questions: 1) Why does it not give me an intercept for the first
> variable and how do i get it; 2) can a get a general estimate of
> correlation, like a wald's statistic for the variable?
1. It's useful to provide a small self-contained example, as recommended
by the posting guide (referred to after every r-help message).
2. I would suggest asking this kind of question on the r-sig-phylo
mailing list, which is devoted to phylogenetic and comparative analyses
in R.
3. In order to understand the output, you have to understand the
way in which R parameterizes statistical models. I'm guessing
that you specified family=binomial in your compar.gee call, which
would mean by default assuming binomial error structure and a
logit link (if you don't know what that means, you should probably
read up on generalized linear models a bit ...). X1 is then
the estimated logit-probability for state 1, X2 is the estimated
*difference* in logit-prob between states 1 and 2, similarly for
X3. If you just want the estimated probabilities for the three
groups you can fit a model without an intercept using something
like "response~predictor-1" as your formula.
4. When you print the model object (e.g. c1=compar.gee(...); c1)
it gives you standard errors and t statistics for each parameter
which are (I believe) essentially Wald statistics, although
you should certainly check the reference given in ?compar.gee ...
good luck,
Ben Bolker
______________________________________________
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.
