There are currently 2 years where market categories for tile are aggregated with each other to form an joint category. They are both done because there isn't enough portside sampling data. The length and age sampling data is used to construct a vector "proportions at length (or age)" for a year. This is then then postmultipled by the "live weight (kg) of commercial landings, to produce "live weight of commercial landings at length (or age)"
There are a few "canned" options. If you exclude a category from the proportions calculation, then you asdume that category is similar to "all" landings. This might be good for unclassifieds or mixed. If you adding a category to a neighboring category then you are assuming the two are identical (say small and extra small).
We want to model this, so we can "pull out" the two aggregation years and predict their class as well.
However, to do this, we need to disaggregate small and kitten. This requires refitting the model.
There are currently 2 years where market categories for tile are aggregated with each other to form an joint category. They are both done because there isn't enough portside sampling data. The length and age sampling data is used to construct a vector "proportions at length (or age)" for a year. This is then then postmultipled by the "live weight (kg) of commercial landings, to produce "live weight of commercial landings at length (or age)"
There are a few "canned" options. If you exclude a category from the proportions calculation, then you asdume that category is similar to "all" landings. This might be good for unclassifieds or mixed. If you adding a category to a neighboring category then you are assuming the two are identical (say small and extra small).
We want to model this, so we can "pull out" the two aggregation years and predict their class as well.
However, to do this, we need to disaggregate small and kitten. This requires refitting the model.