IJMS Editor’s Choice – Unifying statistical modelling through random effects

A selected Editor’s Choice article from the latest issue of the ICES Journal of Marine Science is now freely available. This month read more about the effect of unobserved variables on statistical modelling.
Published: 4 December 2014

​Fisheries biology includes a tremendous and exciting range of research topics. Basic and applied research is increasingly tasked with understanding risks to the environment and the human communities that it supports. These risks are generally assessed using statistical models, which reconcile available data with ecological theory.

In this paper, the authors show that all modelling methods used in fisheries biology have to deal with variables that are observed either indirectly or not at all. Both kinds of variables cause available data to be correlated, which violates the assumptions of simple approaches to statistical modelling. They also show that this difficulty is generally solved by treating unobserved variables as random effects. Random effects therefore unify modelling efforts in the many subfields of fisheries biology.

The authors support this claim by highlighting four examples: the reconstruction of historical changes in population abundance for an exploited fish population, a spatial analysis of fish habitats, an analysis of growth experiments, and an experimental analysis of genetic variation. All four examples show important differences when including or neglecting the effect of unobserved variables. The inclusion of random effects is therefore shown to be a common theme throughout analysis and modelling for fisheries biologists.  

Print this pagePrint it Request newsletterSend to Post to Facebook Post to Twitter Post to LinkedIn Share it
Two wild steelhead trout (Oncorhynchus mykiss) from the American River, on a board used to measure variation in size among indiv

​Two wild steelhead trout from the American River on a board used to measure variation in size among individuals. Individual size is not directly observed except when the fish is caught, so variation in growth must be treated as a random effect; Photo: W. Satterthwaite

Article title: Mixed effects: a unifying framework for statistical modelling in fisheries biology

Authors: James Thorson and Cóilín Minto

c FollowFollow Focus on ContentFocus on Content
HelpGive Feedback

IJMS Editor’s Choice – Unifying statistical modelling through random effects

International Council for the Exploration of the Sea (ICES) · Conseil International pour l'Exploration de la Mer (CIEM)
ICES Secretariat · H. C. Andersens Boulevard 44-46, DK 1553 Copenhagen V, Denmark · Tel: +45 3338 6700 · Fax: +45 3393 4215 ·
Disclaimer Privacy policy · © ICES - All Rights Reserved