Bias in Estimation of Stock-Recruit Function Parameters Caused by Nonrandom Environmental Variability

Abstract

Parameter estimation for stock–recruit models normally assumes a random distribution of residuals around the underlying function. Monte Carlo simulations, in which departures from the mean stock-recruit function were determined by periodic forcing with a random component, showed that bias may occur in the estimation of average parameter values if randomness is assumed. The bias occurred when the spawning stock size varied in-phase or out-of-phase with the periodic forcing and was greatest when the period was approximately twice the mean age of the spawning stock. In addition to bias, patterning of spawner stock size and recruitment data caused by the periodic variability gave misleading impressions of parameter precision.