Using nonlinear hierarchical models for analyzing annulus-based size-at-age data

Abstract

Size-at-age data for fish (derived from otoliths or other structures) are valuable but statistically messy. The data are typically serially correlated and unbalanced, with both time-independent and time-varying covariates. Appropriate growth models are typically nonlinear, with an unknown functional form. We recommend the use of nonlinear hierarchical models for the analysis of such data. We illustrate the use of these methods by applying the recently introduced SAS procedure NLMIXED (SAS Institute Inc., Cary, N.C.) to otolith-based estimated standard lengths of Utah chub (Gila atraria) collected in four locations with predators and four locations without predators.