Set likelihood and penalty weights for the estimation model

Assigns lambda (\(\lambda\)) multipliers to each likelihood component and penalty term in the TMB/RTMB objective function. Weights scale the relative contribution of each data source during estimation and can be used to down-weight noisy data, implement iterative reweighting schemes (e.g., Francis), or disable a component entirely by setting its weight to 1. Must be called after all data setup functions.

Usage

Setup_Mod_Weighting(
  input_list,
  addtocomp = 0.001,
  comp_const_obs = 1,
  addtofishidx = 1e-04,
  addtosrvidx = 1e-04,
  addtotag = 1e-10,
  Wt_Catch = 1,
  Wt_FishIdx = 1,
  Wt_SrvIdx = 1,
  Wt_Catch_pop = 1,
  Wt_FishIdx_pop = 1,
  Wt_SrvIdx_pop = 1,
  Wt_Rec = 1,
  Wt_Init_Rec = NULL,
  Wt_F = 1,
  Wt_Tagging = 1,
  Wt_FishAgeComps = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_fish_fleets)),
  Wt_SrvAgeComps = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_srv_fleets)),
  Wt_FishLenComps = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_fish_fleets)),
  Wt_SrvLenComps = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_srv_fleets)),
  Wt_FishAgeComps_pop = array(1, dim = c(input_list$data$n_pop,
    input_list$data$n_regions, length(input_list$data$years), input_list$data$n_seas,
    input_list$data$n_sexes, input_list$data$n_fish_fleets)),
  Wt_SrvAgeComps_pop = array(1, dim = c(input_list$data$n_pop, input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_srv_fleets)),
  Wt_FishLenComps_pop = array(1, dim = c(input_list$data$n_pop,
    input_list$data$n_regions, length(input_list$data$years), input_list$data$n_seas,
    input_list$data$n_sexes, input_list$data$n_fish_fleets)),
  Wt_SrvLenComps_pop = array(1, dim = c(input_list$data$n_pop, input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_srv_fleets)),
  Wt_Discard = 1,
  Wt_Discard_pop = 1,
  Wt_D = 1,
  Wt_FishAgeComps_discard = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_fish_fleets)),
  Wt_FishLenComps_discard = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, input_list$data$n_sexes,
    input_list$data$n_fish_fleets)),
  Wt_FishAgeComps_discard_pop = array(1, dim = c(input_list$data$n_pop,
    input_list$data$n_regions, length(input_list$data$years), input_list$data$n_seas,
    input_list$data$n_sexes, input_list$data$n_fish_fleets)),
  Wt_FishLenComps_discard_pop = array(1, dim = c(input_list$data$n_pop,
    input_list$data$n_regions, length(input_list$data$years), input_list$data$n_seas,
    input_list$data$n_sexes, input_list$data$n_fish_fleets)),
  Wt_Fish_caal = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, length(input_list$data$lens),
    input_list$data$n_sexes, input_list$data$n_fish_fleets)),
  Wt_Srv_caal = array(1, dim = c(input_list$data$n_regions,
    length(input_list$data$years), input_list$data$n_seas, length(input_list$data$lens),
    input_list$data$n_sexes, input_list$data$n_srv_fleets)),
  fish_sel_pen_wts = NULL,
  ret_sel_pen_wts = NULL,
  srv_sel_pen_wts = NULL
)

Arguments

input_list

Named list with $data, $par, $map, and $verbose sublists, as returned by upstream setup functions.

addtocomp

Small constant added to composition proportions before likelihood evaluation to avoid log(0). Default 1e-3. Ignored when a logistic-normal likelihood is specified, as that family handles zeros internally.

comp_const_obs

Integer switch (0 or 1) controlling where addtocomp is applied in the multinomial likelihood, not a constant to be tuned. 1 (default) adds it to the observed proportions that weight the multinomial as well as inside the logarithms, so the likelihood is stationary exactly at pred = obs. 0 weights by the raw observed proportions. If any fishery or survey conditional age-at-length fleet uses the Dirichlet-Multinomial, 1 triggers a warning, since the added constant biases theta upward when most age bins in a length bin are structurally empty.

addtofishidx

Small constant added to fishery indices. Default 1e-4.

addtosrvidx

Small constant added to survey indices. Default 1e-4.

addtotag

Small constant added to tag recovery observations. Default 1e-10.

Wt_Catch

Weight applied to the catch likelihood. Either a scalar applied uniformly across all fleets, regions, years, and seasons, or a numeric array [n_regions × n_years × n_seas × n_fish_fleets] for fleet- or time-specific weighting. Default 1.

Wt_FishIdx

Weight applied to the fishery index likelihood. Accepts the same scalar or array format as Wt_Catch, dimensioned [n_regions × n_years × n_seas × n_fish_fleets]. Default 1.

Wt_SrvIdx

Weight applied to the survey index likelihood. Accepts the same scalar or array format, dimensioned [n_regions × n_years × n_seas × n_srv_fleets]. Default 1.

Wt_Catch_pop

Weight applied to the population-specific catch likelihood. Either a scalar applied uniformly or a numeric array [n_pop × n_regions × n_years × n_seas × n_fish_fleets]. Default 1.

Wt_FishIdx_pop

Weight applied to the population-specific fishery index likelihood. Same scalar or array format as Wt_Catch_pop, dimensioned [n_pop × n_regions × n_years × n_seas × n_fish_fleets]. Default 1.

Wt_SrvIdx_pop

Weight applied to the population-specific survey index likelihood. Same scalar or array format as Wt_Catch_pop, dimensioned [n_pop × n_regions × n_years × n_seas × n_srv_fleets]. Default 1.

Wt_Rec

Weight applied to the recruitment deviation penalty (ln_RecDevs). Either a scalar applied uniformly or a numeric array [n_pop × n_regions × n_est_rec_devs] for deviation-specific weighting, where n_est_rec_devs is the third dimension of ln_RecDevs rather than the number of years, since dont_est_recdev_last and n_proj_yrs_devs both move it. Default 1. A weight of zero on a deviation leaves it estimated but removes it from the penalty entirely, which is how a stock-recruit relationship is fit over a window of years while recruitment stays free in every year. That is distinct from dont_est_recdev_last, which removes the deviations themselves so recruitment reverts to the deterministic prediction in those years.

Wt_Init_Rec

Weight applied to the initial age deviation penalty (ln_InitDevs). Either a scalar or a numeric array [n_pop × n_regions × (n_ages - 1) × n_sexes]. Defaults to NULL, which takes whatever Wt_Rec is when Wt_Rec is a scalar; supply it explicitly when Wt_Rec is an array, since the two penalties are dimensioned differently.

Wt_F

Scalar weight applied to the fishing mortality deviation penalty (ln_F_devs). Default 1.

Wt_Tagging

Scalar weight applied to the tag-recovery likelihood. Default 1.

Wt_FishAgeComps

Weight applied to the fishery age composition likelihood. Either a scalar or a numeric array [n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_SrvAgeComps

Weight applied to the survey age composition likelihood. Either a scalar or a numeric array [n_regions × n_years × n_seas × n_sexes × n_srv_fleets]. Default: array of 1s.

Wt_FishLenComps

Weight applied to the fishery length composition likelihood. Same format as Wt_FishAgeComps, [n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_SrvLenComps

Weight applied to the survey length composition likelihood. Same format as Wt_SrvAgeComps, [n_regions × n_years × n_seas × n_sexes × n_srv_fleets]. Default: array of 1s.

Wt_FishAgeComps_pop

Weight applied to the population-specific fishery age composition likelihood. Either a scalar or a numeric array [n_pop × n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_SrvAgeComps_pop

Weight applied to the population-specific survey age composition likelihood. Either a scalar or a numeric array [n_pop × n_regions × n_years × n_seas × n_sexes × n_srv_fleets]. Default: array of 1s.

Wt_FishLenComps_pop

Weight applied to the population-specific fishery length composition likelihood. Same format as Wt_FishAgeComps_pop, [n_pop × n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_SrvLenComps_pop

Weight applied to the population-specific survey length composition likelihood. Same format as Wt_SrvAgeComps_pop, [n_pop × n_regions × n_years × n_seas × n_sexes × n_srv_fleets]. Default: array of 1s.

Wt_Discard

Weight applied to the aggregated discard amount or fraction likelihood. Either a scalar applied uniformly or a numeric array [n_regions × n_years × n_seas × n_fish_fleets]. Default 1.

Wt_Discard_pop

Weight applied to the population-specific discard likelihood. Either a scalar or a numeric array [n_pop × n_regions × n_years × n_seas × n_fish_fleets]. Default 1.

Wt_D

Scalar weight applied to the discard mortality rate deviation penalty (logit_dmr_devs). Default 1.

Wt_FishAgeComps_discard

Weight applied to the discard fishery age composition likelihood. Either a scalar or a numeric array [n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_FishLenComps_discard

Weight applied to the discard fishery length composition likelihood. Same format as Wt_FishAgeComps_discard, [n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_FishAgeComps_discard_pop

Weight applied to the population-specific discard fishery age composition likelihood. Either a scalar or a numeric array [n_pop × n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_FishLenComps_discard_pop

Weight applied to the population-specific discard fishery length composition likelihood. Same format as Wt_FishAgeComps_discard_pop, [n_pop × n_regions × n_years × n_seas × n_sexes × n_fish_fleets]. Default: array of 1s.

Wt_Fish_caal

Weight applied to the fishery conditional age-at-length likelihood, multiplying the input sample size of each length bin's age composition. Array [n_regions x n_years x n_seas x n_lens x n_sexes x n_fish_fleets], the shape of ISS_Fish_caal. Defaults to one everywhere.

Wt_Srv_caal

Weight applied to the survey conditional age-at-length likelihood. Same format as Wt_Fish_caal, with n_srv_fleets as the last dimension. Defaults to one everywhere.

fish_sel_pen_wts

NULL (default), or a named numeric vector/list with independent weights for any subset of six selectivity smoothness penalty terms (see resolve_sel_pen_wts and Get_Selex_Smoothness_Penalty), evaluated directly on the fleet's realized selectivity-at-bin-at-year surface and so applicable to any selectivity functional form:

"smooth_bin_curve"

Second-difference (curvature) penalty across bins.

"smooth_bin_diff"

Unconditional first-difference penalty across bins.

"smooth_yr_diff"

First-difference penalty across years.

"smooth_yr_curve"

Second-difference (curvature) penalty across years.

"smooth_dome"

Dome-shape (non-monotonicity) penalty across bins.

"smooth_mean_center"

Per-year mean-centering regularization.

Any name not supplied defaults to 0 (off). Must be called after Setup_Mod_Fishsel_and_Q. Each weight may instead be a vector with one value per model year, so a penalty can act only in some years or with a different strength in each. The specification may also have "bin_range", a length-two vector giving the first and last bin the penalties act over. To give each fleet its own penalties, pass an unnamed list with one named specification per fleet instead of a single specification.

ret_sel_pen_wts

Same format as fish_sel_pen_wts, for the retained fishery selectivity penalty.

srv_sel_pen_wts

Same format as fish_sel_pen_wts, for the survey selectivity penalty. Must be called after Setup_Mod_Srvsel_and_Q.

Value

The input input_list with all weight values stored in $data under their respective names (Wt_Catch, Wt_FishIdx, Wt_SrvIdx, Wt_Catch_pop, Wt_FishIdx_pop, Wt_SrvIdx_pop, Wt_FishAgeComps, Wt_SrvAgeComps, Wt_FishLenComps, Wt_SrvLenComps, Wt_FishAgeComps_pop, Wt_SrvAgeComps_pop, Wt_FishLenComps_pop, Wt_SrvLenComps_pop, Wt_Rec, Wt_F, Wt_Tagging, Wt_Discard, Wt_Discard_pop, Wt_D, Wt_FishAgeComps_discard, Wt_FishLenComps_discard, Wt_FishAgeComps_discard_pop, Wt_FishLenComps_discard_pop).

See also

Other Model Setup: Setup_Mod_Biologicals(), Setup_Mod_Catch_and_F(), Setup_Mod_Dim(), Setup_Mod_Discard_Comps(), Setup_Mod_FishIdx_and_Comps(), Setup_Mod_Fishsel_and_Q(), Setup_Mod_Movement(), Setup_Mod_Rec(), Setup_Mod_Retsel(), Setup_Mod_SrvIdx_and_Comps(), Setup_Mod_Srvsel_and_Q(), Setup_Mod_Tagging()