Evaluation of Assumed Error Structure in Stock Assessment Models That Use Sample Estimates of Age Composition
P.R. Crone and D.B. Sampson
- Price: $0.75 Sale: $0.00
The sampling error associated with estimates of age composition for five groundfish species commercially landed at Oregon ports is used to examine the ability of age-structured stock assessment models to adequately describe the stochastic properties of actual catch-at-age data. Specifically, estimated coefficients of variation associated with samples of catch-atage are presented graphically to evaluate a theoretical consideration involved in stock assessment models widely used in marine fishery management. Results presented here indicate that a multinomial probability error structure, included in models that are based on maximum likelihood estimation, more closely follows the variability associated with the sampled landing data than does a lognormal error structure used in models based on least squares estimation. Weighted nonlinear regression analysis is used to determine the specific multinomial distribution (sample size n) that provides the most accurate description of the actual variability associated with the sample estimates of age composition. Implications for stock assessment modeling are discussed. Finally, a linear regression model is derived that describes the relationship between multinomial sample size and the number of boat trips sampled, in efforts to provide an adequate error structure for the models without having to rely on the relatively complex and tedious sampling estimators and subsequent analytical techniques.
- Item number: AK-SG-98-01q
- Year: 1998
- DOI: https://doi.org/10.4027/fsam.1998.17