To fit sex-specific growth models, the sex of the fish must apparently be known for each observation. However, in some species, the sex of juveniles cannot be visually identified thus generating a bias in the growth fit for early ages. In this study, we address these biases using a latent variable model and testing on several species. The model estimates sex-specific von Bertalanffy growth model using an EM algorithm to include the unclassified sex observations. The species analysed are exploited species having sexually dimorphic growth and partially unclassified observations. Significant differences in parameter estimates were found between the traditional approach of omitting unclassified observations and the EM algorithm. The parameters K and t0t0 of the von Bertalanffy growth model appear to be the most affected (e.g., for herring a 13% increase in the Brody growth coefficient K was estimated). These parameters most impact the fit at young ages. These results show the potential of the model and the implications of omitting unclassified sex observations for fisheries management. Further research is needed to improve and generalise the model, in particular by studying unclassified individuals and applying the EM algorithm to other growth models, while improving the test statistics.
Managing Thesis and Internship Exchanges
