Comparison of Methods for Estimating Mixed Stock Fishery Composition

Abstract

Given information on fish of known origin, and a random sample from the mixed stock fishery, the composition of that mixed fishery may be estimated in a number of ways. This study compares the performance of four classification-based estimators and a maximum likelihood estimator. Theoretical considerations show that the maximum likelihood estimator makes better use of the information contained in the mixed fishery sample. However, the classification estimators are shown to be more robust to violations in some of the model assumptions. Scale data from four regional stock groups of chinook salmon (Oncorhynchus tshawytscha) were used in an applied comparison of the five estimators. The results suggest that the maximum likelihood estimator performs best in practice.