[R] Linear Modeling the best alternative

[Ben Bolker]

stephen sefick <ssefick <at> gmail.com> writes:

> 
> I have a data set of mean velocity, discharge, and mean depth.  I need
> to find out which model best fits them out of log linear, linear, some
> other kind of model...  Using excel I have found that linear is not
> that bad and log10(discharge) vs. the other two variables (I am trying
> to predict velocity and depth from discharge) is not that bad either.
> How do I test and see which one of these models is better...  better
> R-squared...  I know this is a stats question and not particularly an
> R question, but I will use R for the models vetting process.
> any ideas would be greatly appreciated,


  AIC is not bad, but see
http://www.unc.edu/courses/2006spring/ecol/145/001/docs/lectures/lecture18.htm
for computing AIC to compare models where some have transformed
response variables ...

  Ben Bolker