Skip to contents

Prepare length composition data for model input

prep_lf_data.Rd

Transforms a pivoted wide-format length-frequency data frame (one row per fishery x timestep, bins as columns) into the set of model-ready arrays and vectors that opal_model and get_length_like expect. The function validates bin alignment against the length structure in data, optionally filters to a subset of fisheries, and appends every derived object directly to data.

Usage

prep_lf_data(
  data,
  lf_wide,
  lf_keep_fisheries = NULL,
  lf_switch = 1L,
  lf_minbin = NULL,
  lf_maxbin = NULL,
  lf_var_adjust = NULL,
  lf_cap = NULL
)

Arguments

data

a list containing at minimum the length structure scalars len_bin_start, len_bin_width, n_len, and n_fishery.

lf_wide

a data.frame in wide format: columns fishery, year, month, ts plus one numeric column per length bin (named by the bin lower-bound value). Typically produced by tidyr::pivot_wider() on the long-format wcpo_bet_lf object.

lf_keep_fisheries

integer vector of fishery indices to retain. Pass NULL (the default) to keep all fisheries present in lf_wide.

lf_switch

integer likelihood type selector passed straight through to data$lf_switch (1 = multinomial, 2 = Dirichlet, 3 = Dirichlet-multinomial). Default 1L.

lf_minbin

integer vector length n_fishery giving the minimum bin index (1-based) used for each fishery. Defaults to rep(1L, n_fishery).

lf_maxbin

integer vector length n_fishery giving the maximum bin index (1-based) used for each fishery. Defaults to rep(n_len, n_fishery).

lf_var_adjust

numeric vector of length n_fishery used to scale the effective sample size for each fishery. For observation row \(i\) belonging to fishery \(f\), the sample size is adjusted as \(n_i / \text{lf\_var\_adjust}_f\). Values greater than 1 downweight (reduce effective \(N\)) and values less than 1 upweight the fishery. Defaults to rep(1, n_fishery) (no adjustment).

lf_cap

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

Value

The input data list with the following elements appended or updated:

len_lower, len_upper, len_mid

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

lf_switch

Passed through from the argument.

lf_obs_in

Matrix of observed proportions, [n_obs x n_len].

lf_n

Numeric vector of sample sizes per observation row.

lf_fishery

Integer vector of fishery index per observation row.

lf_fishery_f

Integer vector of unique fishery indices with LF data.

lf_n_f

Integer vector of observation counts per fishery.

lf_year

Integer vector of model timestep index (1-based) per observation row.

lf_season

Integer vector of season index (all 1).

lf_minbin, lf_maxbin

Passed through from arguments.

lf_var_adjust

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

removal_switch_f

Integer vector [n_fishery] of removal flags (all 0).

n_lf

Total number of observation rows.

lf_obs_data

Named list list(obs = lf_obs_list) for plain-R access in get_length_like.

lf_obs_flat

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

lf_obs_ints

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

lf_obs_prop

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

lf_nbins

Number of bins used in each observation (scalar, derived from the first fishery's min/max bin setting).

Details

Rows in lf_wide with a total sample size of zero are silently removed before any other processing. Bin alignment between the data frame columns and the model's length structure is checked with stopifnot().

See also

get_length_like, opal_model