Randomization test for the similarity index of communities
We tested similarity indices Cλ and C’λ of communities by randomization tests. Two kinds of null-hypothesis are. assumed in randomization tests of community indices. One of them is that individuals are recorded independently, at random and the other is that samples are taken at random. Randomness of the dispersion of individuals was tested with the dispersion index, m*/m. Individuals were not recorded at random in some field data examined, in which case we did not use the former as a null-hypothesis. In core samples of oribatid mites from five plots with a gradient of vegetation, communities in adjacent plots were not significantly different, but communities of remote plots were significantly different. In repeated transect records of butterflies, both the effects of recorders and time of day were simultaneously tested. The comparison of the similarity index with those of random rearrangement of records is useful for the comparison of communities.