Selecting a Measure of Dispersion

Abstract

In studying the relationship between insect dispersion and population density, we need an index which allows us to separate statistical artifacts from biologically significant effects. I used a simulation model to generate patterns of egg dispersion and tested several dispersion indices as predictors of these patterns. Green's coefficient and standardized Morisita's coefficient were not influenced by population density and are good measures of dispersion. Variance/mean was only weakly correlated with density and has the advantage of being easy to compute and readily understandable. Dispersion indices related to k of the negative binomial are not appropriate for data either more or less clumped than the negative binomial and should only be used with caution.