search

PIFSCstockassessments / AmSam-Bottomfish-2023

Stock assessment of BMUS for American Samoa, scheduled for 2023.
Other
1 stars 0 forks source link

Hindcasting for mean length #47

Closed marcnadon closed 1 year ago

marcnadon commented 1 year ago

This figure appears in the diag report for APVI 50_Base, for example, but not in APRU 50_Base model.

Not clear why this figure appears sometimes and but not always. Felipe wanted to see those figures. Can you look into this when you have the chance Meg?

MOshima-PIFSC commented 1 year ago

@marcnadon I tracked down what is happening. The reason it works for some and not others is based on what years length comp data is available for. APRU only has length data to 2019 (model endyr is 2021) but APVI has it through 2020 (model endyr is 2021). When you only run 3 year peels for the retrospective, there is not enough data to do hindcasting if the length comps end in 2019.
When I tried running SSplotHCxval(retrocomp, subplots = "len") for APRU_50 Base I got the message No observations in evaluation years to compute prediction residuals for Index FISHERY which led me to compare end years of length comp data. If we want to see the hindcast for all length comps, we can increase retrospectives to 4 or 5 year peels (with 4 year peels only have 1 year to use for evaluation so I would recommend using at least 5 year peels).

marcnadon commented 1 year ago

Ok, closing this issue. Thanks Meg