Welcome to the BCSS tutorial on using JAGS and R to analyse wildlife data. We are just starting to put this together so please bear with us where it is incomplete.
The key reason for adopting Bayesian analysis for wildlife data analysis is that it provides useful information for management of wildlife and protected areas. Bayesian posterior probabilities can be used directly in formal decision-making methods. In contrast, frequentist methods based on p-values and null hypothesis significance testing (NHST) can give misleading information.
See Wade (2000) Bayesian methods in conservation biology. Conservation Biology, 14, 1308-1316 for an application to management of beluga whales in Alaska.
JAGS – together with WinBUGS, OpenBUGS and
nimble – uses the BUGS language to code models. This allows biologists to write model code in a way which reflects their understanding of the mechanisms involved.
JAGS works across platforms – Windows, Mac and Linux – and is open source (written in C) and actively maintained. WinBUGS and OpenBUGS do not work on all platforms. They are built on proprietary software which is no longer maintained.
nimble is a relatively new development; we do not include it here (yet!) as we have little experience using it.
Stan is another new package for Bayesian analysis, but it does not use the BUGS language and cannot easily handle the binary latent variables which are so common in wildlife models (eg, present/absent, dead/alive, captured/not captured).
Bayesian analysis is used in a very wide variety of fields from image recognition to genetics. This tutorial focuses on topics relevant to management of animal populations: where species occur, the density or abundance of a population, survival, and community composition.
These share one key feature: our data are imperfect as we do not detect every animal. We need to model the detection process as well as the biological process we are interested in, leading inevitably to hierarchical models.
Author, acknowledgements, IPR
Most of the text for this tutorial was written by Mike Meredith, but incorporates many suggestions and comments from instructors and participants on our real-life “Bayes with JAGS” workshops. As with other BCSS materials, this is made available under a CC-BY-4.0 licence.