search

pfmc-assessments / lingcod

https://pfmc-assessments.github.io/lingcod/
2 stars 2 forks source link

STAR Panel Request 9 #146

Closed brianlangseth-NOAA closed 3 years ago

brianlangseth-NOAA commented 3 years ago
  1. Conduct additional tuning and exploration of the southern model without the 1959-1972 length composition data to see if a reasonable new base model can be developed. Provide the model comparisons slides and table(s) with likelihoods and key parameter outputs, selectivity outputs, etc.

Rationale: Conduct additional tuning and exploration of the southern model without the 1959-1972 length composition data to see if a reasonable new base model can be developed.

brianlangseth-NOAA commented 3 years ago

@kellijohnson-NOAA and @iantaylor-NOAA Im not so sure this is ready for profiles because selectivity parameter 3 for the triennial is still hitting its lower bound. We could certainly do them, but I think it will be more beneficial to resolve the bound issue first (though profiles could inform that) if indeed it is a problem.

Other dynamics are reasonable (Male M = 0.26, Female M = 0.22, h = 0.6) and the scale is now more or less the same as the original base.

You will notice I added the untuned comparisons in 26b635f (even though during the panel I suggested they wouldn't be needed) because the scale is quite different.

Plots are in figures/STAR_request9. I still need to do the table

This is it for me tonight.

iantaylor-NOAA commented 3 years ago

Approach used for the north model was to fix the selectivity parameter 3 at the lower bound: https://github.com/iantaylor-NOAA/Lingcod_2021/blob/26b635f979b1cc28613f4cf13a0396166bdaded2/doc/model-parameters.Rmd#L86-L88 That may not be necessary for good convergence in this case for the south, but I think it's justifiable if we choose to do it.

kellijohnson-NOAA commented 3 years ago

I was able to fix selectivity by adding extra sd to the Triennial index.