Closed kellijohnson-NOAA closed 3 years ago
South model has a the Triennial selectivity parameter on the bound -1. Any thoughts of how quickly we could get the younger age data into the model?
The ageCAAL_S_WCGBTS
and ageCAAL_S_Triennial
available in the package is only sexed fish, so it would involve going back to the data processing.
@brianlangseth-NOAA would know better whether the existing tools could be used or if we should just run some bespoke code for binning the ages associated with small fish.
Alternative approaches to dealing with this parameter on bounds (presumably the descending slope) could be
The last option may pull up the descending limb of other selectivities, including triennial, because old fish would be estimated as dying faster rather than just less observed due to dome-shaped selectivity. There are competing schools of thought: one argues that we should expect dome-shaped selectivity in general for any fishery (see Sampson papers), the other is that if you don't make sure at least one selectivity is asymptotic then you risk having lots of cryptic biomass so it's good to pick at least one fleet to be asymptotic.
Size_DblN_ascend_se_6_Surv_TRI(6)
That's a different story. In that case, I would either lower lower bound and/or free up parameter 5 which allows scaling of the initial selectivity. That is, change the INIT from -999 to maybe 0 and PHASE from -2 to 2 in line 230: -5.00 9 -999.0000 0 0 0 -2 0 0 0 0 0 0 0 #_SizeSel_P_5_6_Surv_TRI(6)
Okay, did both of those things, running now.
@kellijohnson-NOAA Here is the first pass on raw numbers of unsexed fish in the WCGBTS and Triennial surveys. Code would apply for survey_acomps.R. Top rows are ages, bottom and number at age.
As you can see the matter is irrelevant for the Triennial survey (hardly any unsexed aged fish). Most unsexed fish with ages are age 0 in WCGBTS south.
WCGBTS
> table(bioages_north[!is.na(bioages_north$Age) & bioages_north$Sex=="U",]$Age)
0 1 2 3 4 5 6 7
21 13 5 5 7 5 2 2
> table(bioages_south[!is.na(bioages_south$Age) & bioages_south$Sex=="U",]$Age)
0 1 2 3 5
469 29 4 1 1
Triennial
> table(bioages_north[!is.na(bioages_north$Age) & bioages_north$Sex=="U",]$Age)
0 3 4
1 2 1
> table(bioages_south[!is.na(bioages_south$Age) & bioages_south$Sex=="U",]$Age)
0 1
2 7
Plot of length at age for the WCGBTS for these samples
@kellijohnson-NOAA Replying to your question above "Any thoughts of how quickly we could get the younger age data into the model?"
Unsexed fish are partitioned by gender (M or F) for marginal ages, and for lengths for triennial and WCGBTS. For CAAL, the nwfscSurvey package does not have capability to do CAAL for unsexed fish, and we made the call to exclude unsexed fish for CAAL early on. As you mentioned, a hack would be to assign all unsexed fish to female and rerun the CAAL, or to adjust create_caal_nonsurvey.R to include unsexed fish (though I haven't tried to see if that works yet).
Age-7 vs Age-8 was very different for South
Not different for the North
@kellijohnson-NOAA I can revise create_caal_nonsurvey.R to produce CAAL for unsexed fish. I ran this for WCGBTS north and south as a test and it appears to be working. Do you want these still? Given todays direction of discussion, I don't think its needed though, other than perhaps future functionality.
Thanks @brianlangseth-NOAA for looking into the values and @iantaylor-NOAA for the explanation on why they will not matter. I think this issue is done!
Request:
Perform runs for both North and South base models with the Lorenzen mortality function. The reference age for the naturality mortality prior should be age eight (8). Provide the model comparisons slides and tables/figures with likelihoods and key parameter outputs, selectivity outputs, etc. for panel review.
Rationale:
Allowing for a variable natural mortality with age for fish less than age 8 would be a more realistic estimate of natural morality for smaller and immature fish. It is well documented that natural mortality tends to be negatively correlated with body size.