Hierarchical Modeling: A Powerful Framework for Fisheries Research

Fisheries problems become increasingly complicated and fisheries data may contain multiple sources of uncertainty and multiple levels of interdependent structures. We need a realistic modeling framework that captures the complexity of the modern fisheries problems. Hierarchical models, i.e., statistical models of observed data as well as unknown quantities (parameters, latent/hidden variables), are such a modeling framework that can capture uncertainty in data at more than one level. Using several pieces of published and on-going research work as examples, I will demonstrate how to use Bayesian hierarchical models (including multilevel models, mixed effect models, and state-space models) to deal with fisheries data that are subject to the effects of multiple data problems and uncertainty (i.e., sparseness, observation errors and process variation, spatial / temporal / structural interdependence).