2020 ESA Annual Meeting (August 3 - 6)

PS 63 Abstract - Implementing common hierarchical statistical models in ecology with nimbleEcology package for R

Benjamin R. Goldstein1, Daniel Turek2, Lauren Ponisio3 and Perry de Valpine1, (1)Environmental Science, Policy, and Management, University of California - Berkeley, Berkeley, CA, (2)Mathematics and Statistics, Williams College, Williamstown, MA, (3)Entomology, University of California, Riverside, CA
Background/Question/Methods

Certain hierarchical statistical models, such as occupancy, capture-recapture, and N-mixture models, appear in many areas of fish and wildlife biology. The software tool NIMBLE (Numerical Inference for statistical Models using Bayesian and Likelihood Estimation; R package nimble) expanded on widely used tools in the BUGS modeling language family to facilitate the application of these models. NIMBLE offers greater customization of models, algorithms, and Markov Chain Monte Carlo samplers; allows user-defined functions and distributions; and offers improved efficiency through model-specific compilation in C++. In addition to MCMC Bayesian evaluation, NIMBLE supports methods for model selection and validation and methods for maximum likelihood estimation.

Results/Conclusions

We developed nimbleEcology, a supplemental NIMBLE package that provides a suite of useful distributions for ecologists. The goal of nimbleEcology is to make taking advantage of NIMBLE’s capabilities easier by providing ready-made distributions for common ecological models. Rather than independently develop similar distributions, ecologists are able to tap directly into a bank of relevant distributions during model development. nimbleEcology includes distributions for occupancy, dynamic occupancy, capture-recapture, and hidden Markov models, with more planned for the future. Distributions are created using NIMBLE’s nimbleFunctions programming system, making it easy for users to extend them for new purposes.