Influence of Replicated Subsamples and Subsample Heterogeneity on the Linear Index of Food Selection

Abstract

In descriptive studies of prey preference, relative abundances of prey often are calculated from a series of replicated subsamples and total sample sizes are estimated rather than counted. When this is done, the apparent statistical power of the linear index of food selection (Li = ri ‐ pi, where ri and piare proportions of prey taxon i in the gut and habitat, respectively) is artificially inflated. Additional biological variability, due either to environmental heterogeneity or to differences in feeding behavior among individual predators, will inflate further the apparent power of the index and also will produce underestimates of the degree of feeding selectivity. To correct for these problems, expressions that allow for subsampling designs and biological heterogeneity are provided for the standard error of the linear index and the associated degrees of freedom. The ratio of observed sampling variance in relative abundance to that expected for random sampling allows measurement of heterogeneity and statistical testing of the null hypothesis of homogeneity.