For each species, we calculated the D. for each
of the three possible pairwise combinations of forest types.
The area was calculated as the integral of the absolute value of the difference
between distributions; if the ROAPs crossed once, the two sub-areas
were added (so D. was always positive).
Finally, we compared theempirical D. to the distribution generated by randomization and
considered results significant for p < 0.05 using a one-sided test
(i.e., a high observed D.).
We applied a correction for multiple tests,
controlling for false discovery rate by setting a threshold of 0.05 (61
total tests: 13 species for which we performed one habitat pair
comparison plus 16 species for which we conducted three habitat
pair comparisons; Benjamini and Hochberg, 1995).