Hierarchical Bayesian Methods in Ecology (10w2170)

Objectives

Despite the ideological conflict between Bayesian and frequentist systems of statistical reasoning, modern ecologists need to be facile in both methodologies since Bayesian statistics are increasingly prevalent in high ranking ecological journals (Such as Ecology, Ecological Applications and Journal of Applied Ecology). Therefore the objectives of this workshop are:

1. To increase the local scientific community's awareness of Bayesian statistical methods and capabilities.

2. To enable ecologists to understand and customize Bayesian methods.

3. To engage with real datasets of the participants in order to allow theory to be applied to realistic ecological problems

4. To encourage dialogue between ecologists and statisticians.

Although Bayesian theories are not new, their application requires Markov Chain Monte Carlo (MCMC) sampling. The computational power that enables convenient MCMC sampling on personal computers is recent. In the social sciences, researchers have employed Bayesian statistics since the 1990s. Environmental scientists have been reluctant to adopt Bayesian modeling. However In a recent (April 2009) forum published in Ecological Applications, many prominent statisticians and ecologists underlined the suitability of Bayesian methods for hierarchical ecological systems. For complex ecological models, Bayesian methods provide a feasible framework for parameter estimation and accurate inference. Ecological questions typically transcend spatial and temporal scales and therefore require complex hierarchical analysis. The versatility of Bayesian methods endears them to researchers hoping to model complexity and uncertainty at multiple scales. The Bayesian paradigm allows researchers to generate posterior density functions (PDFs) from their data for parameters in ecosystem models. These posterior densities are not limited to estimating linear model coefficients. Modellers can also generate PDFs that reflect the measurement and sampling uncertainty at different times, locations, or levels within the model system. Ecological models should not remain ignorant of the complexity and multiple sources of uncertainty in the environments we study.