fishStan: Hierarchical Bayesian models for fisheries

Abstract

Fisheries managers and ecologists use statistical models to estimate population-level relations and demographic rates (e.g., length-maturity curves, growth curves, and mortality rates). These relations and rates provide insight into populations and inputs for other models. For example, growth curves may vary across lakes showing fish populations differ due to management actions or underlying environmental conditions. A fisheries manager could use this information to set lake-specific harvest limits or an ecologist could use this information to test scientific hypotheses about fish populations. The above example also demonstrates how populations exist within hierarchical structures where sub-populations may be nested within a meta-population. More generally, these hierarchical structures may be both biological (e.g., different lakes or river pools) and statistical (e.g., correlated error structures). Currently, limited options exist for fitting these hierarchical models and people seeking to use them often must program their own implementations. Furthermore, many fisheries managers and researchers may not have Bayesian programming skills, but many can use interactive languages such as R. Additionally, programs such as JAGS often require long run times (e.g., hours if not days) to fit hierarchical models and programs such as Stan can be more difficult to program because it is a compiled language. We created fishStan to share hierarchical models for fisheries and ecology in an easy-to-use R package.