A New Estimator of, and Factors Influencing, the Sampling Variance of the Linear Index of Food Selection

Abstract

Expressions were developed for a new estimator of the sampling variance of the linear index of food selection, Li. This estimator accommodates multiple samples of predators and of their environmental food supply, and can account for variability due to environmental heterogeneity and differences in feeding behavior among predators, if the number of prey counted in a sample is uncorrelated with the proportion of any prey type in that sample. Both laboratory and field studies of young yellow perch Perca flavescens feeding on zooplankton were used to identify factors influencing the estimated variance of Li. The estimated variance decreased with increasing number of fish sampled and increased with increasing observed variance among stomach contents of fish. A heterogeneity index Hri showed significant diet heterogeneity among fish in both laboratory and field experiments. Relationships between the estimated variance of Li and both the number of fish sampled and the total number of prey counted indicated that uncertainty associated with the degree of selectivity of these fish can be reduced to a satisfactory level if samples include at least 10–15 fish and 700–1,500 total prey. Received October 14, 1983 Accepted November 11, 1984