Abstract
Fisheries scientists use different Bayesian software packages to estimate population parameters. Here, four Bayesian algorithms, SIR, Stan, JAGS, and JABBA, were implemented in R to fit a surplus production model with a historical catch and CPUE dataset, five simulated informative catch and CPUE datasets, and five simulated uninformative catch and CPUE datasets. Both informative and uninformative priors applied to the simulated datasets exemplify that informative priors caused the Bayesian algorithms to produce better population parameter estimates, as expected. Specifically, all means and standard deviations of population parameter estimates, except the error terms, either nearly or totally encompassed the actual population parameter values and all means and standard deviations were consistent amongst the algorithms. Analyses with uninformative priors demonstrate that Stan and JAGS are more consistent algorithms and estimate similar means and standard deviations of population parameters that nearly encompass all actual population parameter values, that JABBA estimates encompass nearly all actual population parameters, and SIR does not converge with uninformative priors.