noaa-fims / m1-prototypes Goto Github PK

I am not sure if this is the best place to post this issue. So feel free to direct it somewhere else. I am thinking specifically about Pacific Hake at the moment where we have data back to 1975 but we never "touch" data older than the early 2000s. We do not even have access to the old age comps beyond the Stock Synthesis data file. If you are using season or month what would you do with the older data that doesn't necessarily have a season or month attached to it? Do you always assume that these data are collected in the middle of the year? Is there a better way to denote that the data represent an average over a given time period? I am really torn on how much "pre-processing" of the data should occur before fitting to the model and what stage of the processing should be done within the model versus outside of the model. This is a slightly different topic but related because pre-processing often leads to lack of granularity that you might want later.