User menu
Different approaches were examined in the YFT Assessment examined in 2018, however, conflicting data inputs, data weighting and catchability changes may create problems in the assessment. Issues of convergence/local minima need to be checked thoroughly with jitters. Other diagnostics such as profile likelihood techniques and retrospectives are important to examine. Overall in general this topic needs more time and coverage in the future, as currently the diagnostics examined were limited. In addition, length frequency data are particularly important for the age structured assessments used, and accuracy in these data is crucial to the inference, as large uncertainties still exist in this series (particular attention to ESS and how they influence inference are important). Data weighting issues were not examined extensively, and further work is warranted regarding this subject, as weights between length composition data, CPUE and tagging data can provide very different inferences on the population. Overall, the process was transparent, and issues were briefly discussed relevant to uncertainty in the assessment results. A key limitation was that insufficient time was available to examine both data and assessment issues at the meeting. If we could discuss model resolution and data before the meeting, additional time would be available to discuss further refinements in the assessments. Preliminary analysis using hindcasting techniques suggest that the model has poor predictive power which is a concern, and may also be problematic as it has local minima issues. Finally, approaches dealing with uncertainty and projections were not given due importance, but as these are critical for stock status advice, and management advice that would sustain the long-term sustainability of the stock, additional time should be spent on these issues in the future.
