Tau as an index of similarity in community comparisons: an approach permitting the hypothesis of unequal species abundance

Abstract

P. A. Jumars (Ecology, 61: 1553–1554. 1980) has argued that in significance tests of τ where it is employed in community comparisons, the null hypothesis that all rankings of species are equally likely can be translated only into the hypothesis that all species are equally abundant. An alternative translation is presented, in which it is the species names that are equally likely to be attached to unequal partitionings of the population. Some biological justifications for this alternative model are offered. It is shown that Jumars' example calculations illustrate type II rather than type I statistical errors, so that the "serious doubt" that he expresses about published community comparison significance tests applies to acceptances rather than rejections of the null. The alternative index of community similarity he proposes, based upon contingency table analyses of species counts, is shown to reflect excessive sampling homogeneity rather than community similarity.