Using highest density intervals can reduce perceived uncertainty in stock assessments

Abstract

Sustainably managed fisheries provide conservation and socio-economic benefits. To set catch quotas, decision makers are guided by stock assessments that use mathematical models to make quantitative predictions of how populations will change under different management scenarios. Stock assessments need to communicate uncertainty of estimated quantities, which is often done through figures and tables depicting credible or confidence intervals. We show that computing such intervals with the usual equal-tailed approach has undesirable consequences, such as excluding highly probable values yet including relatively improbable values. This can give an overly optimistic impression of the health of fish stocks, with potential unexpected management implications. Instead, using highest density intervals can resolve these problems, resulting in narrower intervals that reduce perceived uncertainty (by $>$3 billion fish for a recent cohort of Pacific Hake, Merluccius productus, for example). Therefore, we recommend consideration of highest density intervals in fisheries applications and other fields to better characterize uncertainty and improve conservation advice. We introduce our new R package, hdiAnalysis, which enhances previous methods to encourage and facilitate uptake by practitioners.