Closed gavinfay closed 3 years ago
Thanks, I was wondering what to do with these. They are 0s in the Atlantis YOY.txt file and I have been trying to track down the reason because they should have recruitment. There is also -Inf in the average recruitments for these species. So I may have to dig into the .nc file.
I can fill the devs with 0s. Based on our conversation. I will also pull the age 1s and get the average of those over the modeled time period for the -Inf values
@gavinfay new .pin file now includes Atlantis YOY based average recruitment and devs for everyone who has it, and subs in average age 1 N for average recruitment and 0 for recruitment devs for those who don't (2 species).
the whole mess is here https://github.com/ices-eg/wg_WGSAM/blob/d54a082c7fe5b0795de299eeb92283e910b46975/SkillAssessment/SkillAssessProject.Rmd#L1293-L1357
If this level of (in)consistency seems too strange even for starting values, I could use the average age1 N and deviations for everyone. However, these match roughly in pattern, but not in magnitude for the species that have both. I suppose it makes sense there is still substantial mortality between "recruitment" in Atlantis and the within year timestep with max numbers at age 1.
new pin file uploaded with https://github.com/thefaylab/hydra_sim/commit/9762838017783656f135572ac111afcd2f631f67
Species 3 & 9 have
-999
as vectors of therecruitment_devs
array in the NOBA pin file. @sgaichas I am assuming this is because these species have a different specification for recruitment, but we will need to find a different solution as this array is an estimated parameter (and needs to be differentiable). If the intent is to model no deviations from the stock-recruit relationship for these species we can set to 0 and change phase to a negative number.We can set different estimation phases for different elements of a
init_bounded_number_vector
but I do not remember / am not sure (yet) if this is achievable in the same way for a 3darray.