Workshop on Statistical Methods for Dynamic
Workshop on Statistical Methods for Dynamic
Vancouver, June 4-6 2009
System Models
Fisheries scientists working in the field of fish population dynamics provide quantitative advice to fisheries managers about how alternative harvesting regulations might affect their ability to achieve management objectives such as maximizing catch or producing a small chance of collapsing a fish population. To generate that advice, scientists fit models of various complexities to field data, but the statistical properties and long-term consequences of choosing particular estimators are usually unknown because these models tend to be nonlinear with thresholds and delays of varying length. It is therefore critical that fisheries scientists evaluate the performance of candidate estimators in computer simulation experiments. For example, fisheries scientists often build empirically-based, stochastic "operating models" of fish population dynamics to generate "observed" data from user-specified underlying "true" parameter values and functions. Simulated data can then be analysed by using a range of candidate estimators and then comparing the resulting parameter estimates with "true" values. Often, however, the "best" parameter estimation methods (i.e., those that best achieve management objectives) are not necessarily those with the best statistical properties (bias, precision, AIC value, etc.) because fisheries systems generally have asymmetric rather than quadratic loss functions. Therefore, some fisheries scientists go further and simulate entire non-linear feedback systems by combining empirically-based models for dynamics of the natural ecological system, parameter estimation of models fit to simulated data, and setting of harvest regulations by simulated decision makers using algorithms based on those parameter estimates. We will show several examples of empirically-based simulation analyses of non-feedback and feedback fishery system models.
Development and testing of parameter estimation methods for modelling fisheries systems
Professor and Canada Research Chair in Fisheries Risk Assessment and Management
School of Resource and Environmental Management
Simon Fraser University
Burnaby, British Columbia
Assistant Professor
School of Resource and Environmental Management
Simon Fraser University
Burnaby, British Columbia