Fishing mortality, discard mortality and selectivity deviations in the operating model

Stores what each replicate needs to draw these deviations and rebuild the fleet arrays from them, and draw_sim_fleet_devs does so in Setup_sim_env, keeping the fit's deviations over the simulation list's n_cond_yrs and drawing the fitted years after them.

Usage

Setup_Sim_Fleet_Devs(
  sim_list,
  data,
  pars,
  random = NULL,
  pars_by_sim = NULL,
  F_devs_draw = c("random", "all")
)

Arguments

sim_list

Simulation list with n_yrs, n_sims, Fmort, dmr and the selectivity arrays already set.

data

Data list of the fit.

pars

Parameter list at the fitted values.

random

Character vector of the fit's random effects. Default NULL.

pars_by_sim

Optional list of one parameter list per replicate, for replicates that each run on their own parameter draw (sim_type = "joint" in simulation_self_test). NULL (default) gives every replicate pars.

F_devs_draw

"random" (default) draws F and discard mortality deviations only when random integrates them out; "all" also draws them when the fit holds them as fixed effects, from their penalty.

Value

sim_list with fleet_devs_on (which of F, dmr, fish, ret and srv are drawn), fleet_dev_data and fleet_dev_pars added, or unchanged when nothing is drawn.

Details

Selectivity deviations are drawn wherever the fit varies selectivity, under the process Get_PE_loglik penalizes, as growth and movement deviations are. F and discard mortality deviations are drawn by default only when the fit integrates them out (random): as fixed effects their penalty is usually a loose constraint, sigmaF = 1 by default, and drawing from it would replace the fitted F with noise. F_devs_draw = "all" draws them as fixed effects too, from the same penalty. Either way they are kept at the fit when the penalty is off or centered on the deviations' own mean, neither of which is a density to draw from.

A penalty weighted by Wt_F, Wt_D or *_pe_wt is the density of the same process with its variance divided by the weight, and that is what is drawn; a self test keeps the weight in its refits.

Selectivity at length (*_selex_type = 1) is drawn and rebuilt at length, and its selectivity at age is the curve read through each replicate's own size-age transition matrix, by the growth rebuild when the operating model rebuilds growth and directly otherwise. A penalty weight w (*_pe_wt) scales the penalty to the density of a process whose variance is divided by w, which is what is drawn. When the operating model runs past the fit, as a closed loop does, selectivity deviations are drawn in those years too, on from the fitted ones (see draw_sim_fleet_devs). F and discard mortality never are, since a closed loop sets F in its projection years by its control rule.

See also

Other Simulation Setup: Setup_Sim_Biologicals(), Setup_Sim_Containers(), Setup_Sim_Dim(), Setup_Sim_Fishing(), Setup_Sim_NAA_state(), Setup_Sim_Rec(), Setup_Sim_Survey(), Setup_Sim_Tagging(), Setup_sim_env(), Simulate_Pop_Static(), run_annual_cycle(), simulation_self_test()