We used non-parametric analytical methods; therefore results
presented are for untransformed data. However, transformation
can decrease the variance around the mean (Sokal and Rohlf,
1995) potentially influencing splitting in permutation trees, thus
we double-checked our findings with log + 1 transformed data
and found no significant differences. We tested all variables for
colinearity. Density of trees and basal area were collinear (pearson
correlation coefficient = 0.501, p = 0.0007), so they were not tested
simultaneously in the analyses. Species accumulation and estimation
curves were generated using the Chao1 estimator (Chao,
1987),