A new electivity index for diet studies that use count data

Abstract

AbstractElectivity indices summarize the results of field‐based feeding studies by comparing the relative abundance of a potential prey item with its relative prevalence in the diet of a predator. We developed a new electivity index based on odds ratios calculated from counts of individual prey taxa in ambient samples and in the gut of a predator. Many indices of electivity have been developed, though the literature lacks consensus on which is the best. Why use our index instead of one of these? Most of the extant indices lack the means for assessing uncertainty, treat proportions determined from count data as fixed rather than estimates, and ignore the skewness inherent in binomial data. The new index is calculated by Bayesian estimation from the binomial distributions of proportions in ambient samples and gut contents, providing full likelihoods of the index and its components. Indices from groups of predators can be aggregated without bias using the log odds ratio. Using simulated data, we show how the credible intervals of the index shrink with increasing numbers of total ambient and consumed prey and become increasingly asymmetrical as the index approaches its limits (0 and 1). Applying the method to diet data for an endangered planktivorous fish, we show how aggregating among fish and among samples was necessary to overcome the limitations imposed by low counts and high variability among individual fish. The indices for the smallest of five prey taxa varied inversely with total prey abundance, consistent with optimal foraging theory.