Feedback and Modularization in a Bayesian Meta-Analysis of Tree Traits Affecting Forest Dynamics

• Activity Number: 168
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kiona Ogle*+ and Jarrett Jay Barber
• Companies: Arizona State University and Arizona State University
• Keywords: incomplete reporting ; feedback control ; modularization ; ecological meta-analyis ; hierarchical Bayesian ; forest ecology

Abstract

We describe a unique application of modularization, or 'feedback control', in the context of a Bayesian meta-analysis of literature information. Numerous missing data are common to meta-analyses, and, in this study, poor chain mixing and identifiability issues resulted. In response, we modularized model components such that missing covariate data do not allow feedback between modules to affect parameters in the covariate module (direct feedback control) or to affect covariate effects parameters in the mean model for the response (indirect feedback control). Our use of direct and indirect feedback control improved mixing and convergence, yielding realistic pseudo-posteriors. Such modularization addresses limitations of existing meta-analytic methods by accommodating incomplete reporting and by considering all quantities as stochastic, including sample means (response), sample sizes, standard errors, and covariates, reported or not. We illustrate our approach with literature information on specific leaf area, a key parameter in models of tree growth and forest dynamics. We discuss problems that arise from feedback between modules and provide ecological arguments for modularization.