A Variance Test for Detecting Species Associations, with Some Example Applications

Abstract

A little—used measure of species association is provided by the ratio of the variance in total species number (or total density of individuals) in samples to the sum of the variances of the individual species. Use of this ratio allows a test of the null hypothesis that species do not covary among samples. As the test is simultaneous it is preferable to pairwise 2 X 2 contingency table analysis and pairwise correlation analysis when the number of species is large. Simulation reveals that this test is also more powerful than Pearson's g 2 test for comparing observed and expected frequency distributions of species in samples. The variance test was applied to a selection of data from the literature. Results reveal that positive associations are the rule in nature and negative associations are uncommon. Invertebrate species tend to be more positively associated among samples than are vertebrates, although the difference is not significant. Examples of biological processes, especially interspecific interactions, that might produce associations among species are suggested. However one must be cautious not to infer the importance or unimportance of a specific process from the result of the test alone.