BUGS in Bayesian stock assessments

Abstract

This paper illustrates the ease with which Bayesian nonlinear state-space models can now be used for practical fisheries stock assessment. Sampling from the joint posterior density is accomplished using Gibbs sampling via BUGS, a freely available software package. By taking advantage of the model representation as a directed acyclic graph, BUGS automates the hitherto tedious calculation of the full conditional posterior distributions. Moreover, the output from BUGS can be read directly into the software CODA for convergence diagnostics and statistical summary. We illustrate the BUGS implementation of a nonlinear nonnormal state-space model using a Schaefer surplus production model as a basic example. This approach extends to other assessment methodologies, including delay difference and age-structured models.