[R-sig-ME] Finding r^2 value and plotting regression lines and CIs from gls

[Ben Bolker]

Margie Mathewson <m1mathew at ...> writes:

> 
> I'm a new R user who has been using the gls function to analyze scaling
> relationships in mammals while considering phylogenetic similarity. I'm
> having trouble with a few things, though.
> 
> 1) For my project, I need to know the r or r^2 value of my scaling line,
> and I'm having trouble figuring out how to find it in R. Is there an easy
> way to get gls function to spit out this value?

   R and R^2 are a little bit complicated once one gets beyond simple
linear models -- you have to decide exactly what you mean by "variance
explained"; you can search for R^2 or pseudo-R^2 on this list or elsewhere
for excruciating details.   In the meantime, methods(class="gls") tells
you what you can do with a gls fit, and the correlation between predicted
and observed values is a reasonable **CRUDE** approximation of R:

example(gls)
cor(predict(fm1),Ovary$follicles)

predict() is also what you need to draw a regression line 
(see ?predict.gls)

> 
> 2) Is there a way to plot the regression line and confidence intervals
> calculated by the program onto the data sets I'm analyzing in R? I've been
> able to plot residuals, but not the regression line and data.

confidence intervals are probably pretty tricky for the aforementioned
reasons ...

> 
> Thanks for any suggestions you can offer!
> Margie
> 
>