Premium
Drawbacks of complex models in frequentist and Bayesian approaches to natural‐resource management
Author(s) -
Adkison Milo D.
Publication year - 2009
Publication title -
ecological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.864
H-Index - 213
eISSN - 1939-5582
pISSN - 1051-0761
DOI - 10.1890/07-1641.1
Subject(s) - frequentist inference , prior probability , computer science , bayesian probability , bioeconomics , natural resource , simple (philosophy) , expert elicitation , natural resource management , bayesian inference , machine learning , artificial intelligence , ecology , mathematics , fishery , statistics , philosophy , epistemology , biology
Previous studies have shown that, for managing harvest of natural resources, overly complex models perform poorly. Decision‐analytic approaches treat uncertainly differently from the maximum‐likelihood approaches these studies employed. By simulation using a simple fisheries model, I show that decision‐analytic approaches to managing harvest also can suffer from using overly complex models. Managers using simpler models can outperform managers using more complex models, even if the more complex models are correct and even if their use allows the incorporation of additional relevant information. Decision‐analytic approaches outperformed maximum‐likelihood approaches in my simulations, even when Bayesian priors were uninformative.