Selecting Loci for Genetic Stock Identification Using Maximum Likelihood, and the Connection with Curvature Methods

Publisher Website
Google Scholar
Abstract
Maximum likelihood theory is used to predict the precision of genetic stock identification composition estimators — prior to collection of the mixed fishery sample. It is shown how this allows the researcher to plan the genetic stock identification study, through specification of sample size and choice of genetic data to assay, so as to maximize estimator precision. The curvature methodology used in Gomulkiewicz et al. (1990. Can. J. Fish. Aquat. Sci. 47: 611–619) is shown to be closely related to the maximum likelihood approach. In that study, interpretation of results is complicated by the use of an overparametrized curvature measure. Here it is shown that when applied to an appropriately parametrized likelihood function the curvature methodology reproduces the maximum likelihood theory.