[R] comparing distributions

[kdebusk]

I have data sets from three different sites. None of the data sets are
normally distributed and can not be log-transformed to a normal
distribution. I would like to compare the data sets to see if any of
the sites are similar to each other, and would like something more
powerful than a non-parametric wilcoxen test. Does anyone have any
suggestions on more powerful ways of comparing non-normal data sets?
If it helps, the data sets follow a Gamma distribution. Thanks for all
input.