Choosing priors in Bayesian ecological models by simulating from the prior predictive distribution

Bayesian data analysis is increasingly used in ecology, but prior specification remains focused on choosing non-informative priors (e.g., flat or vague priors). One barrier to more informative that must be specified model parameters intercepts, slopes, and sigmas), knowledge often exists the level of response variable. This particularly true for common models like generalized linear mixed have a link function potentially dozens parameters, each which needs distribution. We suggest this difficulty can overcome by simulating from predictive distribution visualizing results scale In doing so, some choices easily seen produce biologically impossible values variables. Such implications are difficult foresee without visualization. demonstrate workflow selection using simulation visualization with two ecological examples (predator–prey body sizes spider responses food competition). approach not new, its adoption ecologists will help better incorporate information models, thereby maximizing one benefits analysis.