Correction to 'Bayesian inference reveals positive but subtle effects of experimental fishery closures on marine predator demographics'

Abstract

It has recently come to our attention that we made a coding error while implementing the hierarchical mixed-models describing chick condition and chick survival (equations (2.1) and (2.2), respectively) in ‘Bayesian inference reveals positive but subtle effects of experimental fishery closures on marine predator demographics' [1]. This error meant that the nested random effect described in equation (2.1) and the hierarchical shared frailty term described in equation (2.2) were not implemented correctly in the original analysis. This resulted in our reporting the model parameter estimates with higher precision than should have been the case (table 1 and figure 1). Here we present corrected results (table 1 and figure 1) based on correctly specified models in JAGS [2] applied to the datasets used in our original analysis. Our error, which pertained to the specification of the structure of the priors for the nested random effects (see electronic supplementary material), influenced the derived parameter estimates of the mean chick condition or chick survival during closed and open years (figure 1) more (standard deviations are between 29% and 285% larger in the corrected results) than the estimates for the regression coefficients representing the island effect (–0.8 to 9%), closure effect (4–8%) or their interaction (6–8%). These changes do not alter the predicted influence of the fisheries closures on the population dynamics of endangered African penguins Spheniscus demersus.