Bayesian surplus production models for mako shark, using alternative integration algorithms
The Bayesian Surplus Production (BSP) software, which uses the Sampling-ImportanceResampling (SIR) method to integrate posterior distributions, was used for the ICCAT mako assessments through 2012. The 2014 assessment of blue shark used both the BSP software and the Markov Chain Monte Carlo (MCMC) algorithm, implemented in the JAGS software, and found that the JAGS and BSP model results were not always consistent. We applied both the BSP1 software (without process error) and the BSP2 software (with process error), and two independent MCMC software packages, JAGS and Stan, to the data from the 2012 mako shark assessment for the North Atlantic to determine whether the same problem exists. We also used the SIR and MCMC algorithms from LearnBayes to fit the same function with both algorithms. Although all the modeling approaches give fairly consistent posteriors for r, the posteriors of K were somewhat different. This may be because there is a long period of catches with no CPUE data, or because the catch and CPUE data are not consistent with each other. The lack of information in the data may cause the model to be sensitive to minor differences in how the model is configured.