Dear Jason, Beta regression is what you should be looking at. The beta distribution is appropriate for data on proportions of continuous variables (e.g. vegetation cover). You just need yo convert your percentages to proportions (dividing by ten). Each plot where you measured vegetation cover will be your replicate (or whatever sampling design you made) and will have a value between 0 and 1.
I hope this helps. Andrés Andres Lopez-Sepulcre Laboratoire d'Ecologie, UMR 7625 Ecole Normale Superieure, Paris [email protected] http://www.ecoevo.net/LopezSepulcre On Sep 8, 2013, at 7:19 AM, Jason Hernandez <[email protected]> wrote: > > Dear Ecologgers: > > I have a data set comparing percent cover in two types of habitat. In both > treatments, I set my protocol such that percentages sum to 100 in each > sample, and then I took the averages and made pie charts to illustrate the > difference. The visual difference between the two pie charts is striking: in > one, a single cover class fills 95% of the pie, and there are only 5 cover > classes all together; in the other, no one cover class is more than 45%, and > there are eight cover classes. > > My question: how should I structure these data in order to conduct meaningful > statistical tests? The sample size is (by design) the same in both > treatments. Thanks. > > Jason Hernandez
