Mirror selectivity to one time block from another fleet

[Kiva Oken - NOAA Federal]

Hi,

We have a fishery independent survey with some limited composition data that we were not planning on including, because its selectivity is likely to be very similar to the selectivity of one of the catch fleets that has plenty of data. However, the fleet in the model has time blocks to represent regulatory changes. The survey overlaps those time blocks, but selectivity should be consistent. If the time blocks are applied to the survey index, they would effectively change catchability of the survey, even though the sampling process did not change. 

We would like to use the base selectivity from the catch fleet to be applied across the entire survey duration. Is it possible to do that in stock synthesis? We have not been able to figure it out yet.

Thanks,

Kiva

--

Kiva Oken, PhD (she/her)
Research Mathematical Statistician, Northwest Fisheries Science Center
NOAA Fisheries | U.S. Department of Commerce
(564) 225-2297
kiva...@noaa.gov
www.fisheries.noaa.gov

I live and work on the traditional lands of the Coast Salish peoples.

[Richard Methot - NOAA Federal]

Hi Kiva,

This is a logical situation, so hope we can figure out a solution.  I think what would work best would be if there was a way to mirror some selectivity parameters across fleets/surveys rather than to mirror the resultant selectivity, which is the current approach.  So, this is not an immediate solution.

SS3 updates a fleet's selectivity:

    if (timevary_sel(y, f) == 1 || save_for_report > 0)
    { // recalculate the selex in this year x type

But creation of that flag is aware of mirroring, so we cannot mirror to the base fleet selectivity and ignore mirroring for changes to the fleet's selectivity.

One approach might be to break the fishery fleet into a pre and post-regulatory change fleet then mirror the survey to only the pre-regulation fleet.

Another might be to make the survey the lower numbered fleet and to mirror the base selectivity for the fleet to the survey, but I do not see how to then allow for time-varying parameters for the fleet.

I'll keep thinking.....

Rick

[Jason Cope]

I was also going to suggest breaking the fishery into two blocks in order to isolate the portion that the survey could be mirrored too. Blocks vs separate fleets with different selectivities should be equivalent practices.

A rougher approach would be to estimate the selectivities for the fishery, then fix the survey to the appropriate time block values. 

I prefer the first, but the second likely is also be good enough.

-Jason

[Kiva Oken - NOAA Federal]

Thanks Jason and Rick! I agree, it seems like two fleets is the best option with the current version of stock synthesis. Though I will note it is not actually the same because the two fleets will have different weights for comp data, which they should not. (I suppose you could manually match the weights, but I don't know of a quick and easy way to do that.)

--
You received this message because you are subscribed to the Google Groups "SS3 - Forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ss3-forum+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/04fc9ef6-bc7b-41d1-a7b5-4928cd5bed7dn%40googlegroups.com.

[Ian Taylor - NOAA Federal]

Hi Kiva,

Good point about the composition weights!

The r4ss functions for calculating the Francis method weights (SSMethod.TA1.8 and SSMethod.Cond.TA1.8) include a "fleets" argument which is defined as "vector of one or more fleet numbers whose data are to be analysed simultaneously (the output N multiplier applies to all fleets combined)" but this can't yet be controlled in the convenient wrapper function r4ss::tune_comps(). I think it wouldn't be too hard to add.

The Dirichlet-multinomial option allows repeated values in the ParmSelect column which will lead to shared weighting parameters for the same fleet.

-Ian

To view this discussion visit https://groups.google.com/d/msgid/ss3-forum/CACsG4K7k3dvULBs%3Dz3R4T%2Bv2ATmjc32xwhYnjhuYCOiF3-SQUQ%40mail.gmail.com.

[Mark Maunder]

Hi Everyone,

 

Does SS output a Length-Age Key (i.e. the distribution of age conditioned on length)? Or do I need to make a dummy age-conditioned on length for each time period (and fishery if there is age based selectivity) to get the values?

 

This would allow me to create catch-at-age data from the observed length compositions to use as a diagnostic.

 

Thanks,

 

Mark