Prepare weight composition data for model input

Transforms a pivoted wide-format weight-frequency data frame into model-ready arrays and vectors that get_weight_like() expects. Also precomputes the rebinning matrix for converting predicted length compositions to weight compositions.

Usage

prep_wf_data(
  data,
  wf_wide,
  wf_keep_fisheries = NULL,
  wf_switch = 1L,
  wf_minbin = NULL,
  wf_maxbin = NULL,
  wf_var_adjust = NULL,
  wf_cap = NULL
)

Arguments

data

list containing at minimum: lw_a, lw_b, len_bin_start, len_bin_width, n_len, n_fishery, and the weight bin scalars wt_bin_start, wt_bin_width, n_wt.

wf_wide

data.frame in wide format: columns fishery, year, month, ts plus one numeric column per weight bin (named by bin value).

wf_keep_fisheries

integer vector of fishery indices to retain (NULL = keep all).

wf_switch

integer likelihood type (1=multinomial, 2=Dirichlet, 3=Dirichlet-multinomial). Default 1L.

wf_minbin

integer vector (length n_fishery) min weight bin index. Defaults to rep(1L, n_fishery).

wf_maxbin

integer vector (length n_fishery) max weight bin index. Defaults to rep(n_wt, n_fishery).

wf_var_adjust

numeric vector (length n_fishery) variance adjustment divisors. Defaults to rep(1, n_fishery).

wf_cap

positive integer (or NULL). When supplied, each effective sample size is capped: \(n_i = \min(n_i, \text{wf\_cap})\). Applied after proportions are computed so that observed compositions are unaffected. Default NULL (no cap). This is a Multifan-CL legacy feature.

Value

data list with the following weight composition elements appended:

wf_switch

Passed through from the argument.

wt_lower, wt_upper, wt_mid

Weight bin boundary and midpoint vectors derived from the scalar inputs.

wt_bin_edges

Weight bin boundary vector (length n_wt + 1).

wf_rebin_matrix

Precomputed rebinning matrix (n_wt x n_len) for converting predicted length compositions to weight compositions.

n_wf

Total number of WF observation rows.

wf_obs_in

Matrix of observed proportions (n_wf x n_wt).

wf_obs_flat

Flattened numeric vector of counts (for multinomial, wf_switch = 1).

wf_obs_ints

Flattened integer vector of rounded counts (for Dirichlet-multinomial, wf_switch = 3).

wf_obs_prop

Flattened numeric vector of normalised proportions (for Dirichlet, wf_switch = 2).

wf_n

Numeric vector of sample sizes per observation row.

wf_fishery

Integer vector of fishery index per observation row.

wf_fishery_f

Integer vector of unique fishery indices with WF data.

wf_n_f

Integer vector of observation counts per fishery.

wf_year

Integer vector of model timestep per observation row.

wf_minbin, wf_maxbin

Passed through from arguments.

wf_var_adjust

Passed through from argument; numeric vector [n_fishery] of variance-adjustment divisors applied to wf_n.

See also

prep_lf_data, opal_model