The accuracy of alternative stochastic growth rate estimates for salmon populations

Abstract

The accuracies of four alternative estimators of stochastic growth rate for salmon populations are examined using bootstrapping. The first estimator is based on a stochastic Leslie matrix model that uses age-specific spawner counts. The other three estimators use spawner counts with limited age-structure information: a Botsford–Brittnacher model method and two diffusion approximation methods, namely, the least squares approach of Dennis and the robust approach of Holmes. Accuracy of the estimators was quantified using median bias and interquartile ranges of the stochastic growth rate estimates. The Botsford–Brittnacher estimator was found to be unreliable due to large bias. Of the remaining estimators, the stochastic Leslie approach tended to produce the most reliable estimates but had the greatest data demands. With severe lognormal measurement error, the Dennis estimators produced less biased estimates than the other methods, but precision of the stochastic growth rate was generally highest using the stochastic Leslie estimator.