Is down-weighting composition data adequate for dealing with model misspecification, or do we need to fix the model?

Abstract

Data weighting has become a major area of focus for fisheries stock assessment research. The sensitivity of results to data weighting is due mainly to model misspecification. We develop an age-structured population dynamics model for summer flounder (Paralichthys dentatus) in the U.S.A. mid-Atlantic region to explore the influence of model misspecification and data weighting on estimation performance using simulation. Our results confirm those of numerous previous studies that model misspecification can lead to substantial bias in estimates of quantities of management interest. We also show that data weighting generally cannot counter the effects of model misspecification, except by a small amount when composition data from a fishery for which the selectivity is misspecified are down-weighted. Additionally, estimating natural mortality was not useful as a way to address model misspecification, except when natural mortality itself was misspecified. In conclusion, down-weighting composition data is generally not adequate for dealing with model misspecification. Therefore, we recommend that more work be done to ensure that models are correctly specified. Unfortunately, there is still a lack of understanding and uncertainty in the fundamental biological and fishing processes of most, if not all, fish stocks, making the removal of model misspecification difficult.