Variance Partitioning and Nongeographic Variation

Abstract

The variability of characters within populations is generally measured as a function of the character variance, although questions of nongeographic variation are addressed as tests of mean differences between subgroups within populations. Presented here is an alternative approach based upon partitioning the variance of a character into components due to the presence of such subgroups within the population. The relative magnitude of these variance components is used to determine the relative importance of sources of variation within populations and to examine patterns of variability across taxonomic groups. The method is flexible enough to utilize data with unequal sample sizes and provides estimates which are directly comparable across characters and taxa.