Estimating Variability in Biodiversity--Ecosystem Function Experiments with Resampling Methods

• Activity Number: 275
• Type: Contributed
• Date/Time: Tuesday, August 10, 2004 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economics Statistics Section
• Author(s): Kevin Gross*+ and Bradley J. Cardinale
• Companies: North Carolina State University and University of Wisconsin, Madison
• Keywords: biodiversity ; ecology ; bootstrap

Abstract

In ecology, several high-profile experiments have been conducted recently to elucidate the effects of biodiversity loss on ecosystem function. In these experiments, communities are assembled by random selection of species from an experimental species pool, and relationships between the number of species in a community and ecosystem function are sought. From a statistical perspective, shared dependence of the experimental communities on the species pool induces strong correlations among community responses. To generalize the results of these experiments, it is necessary to understand how this correlation structure affects the sampling variability in the observed biodiversity--ecosystem function patterns. Here, we propose three resampling methods (bootstrap, jackknife, and infinitesimal jackknife) to estimate this sampling variability, and study their performance in simulations. Our results suggest that these basic re-sampling methods tend to overestimate sampling variability, and that opportunity remains for the development of a superior resampling method.