Locate every estimated parameter in the model's own coordinates

One row per estimated (fixed-effect) parameter, giving the fleet, species, sex, age, bin, year or slot it occupies.

Usage

parameter_index(object)

Arguments

object

A fitted Rceattle model from fit_mod().

Value

A data frame with one row per estimated parameter and columns par_index (position in obj$par), block, species, fleet, sex, age, bin, year, slot, n_cells and label.

Details

TMB names each element of obj$par after its parameter block, so a convergence diagnostic can report that sel_coff_dev is non-identifiable but not which of its elements.

The mapping is recovered by pushing a tagged parameter vector back through TMB's own parList(), so it reflects build_map() exactly: parameters mapped off are absent, and fleets sharing a Selectivity_index or Catchability_index appear once, as the single parameter they are, with n_cells recording how many array cells they drive.

Ages are absolute. init_dev holds one deviation per age from minage + 1 upward, since age minage in the first year is recruitment.

Coordinate columns are NA where the axis does not apply to a block. The linkage and environmental-covariate blocks (beta_linkage, M1_beta, ...) are indexed by a linkage-table row rather than a model coordinate, so they hold no coordinates and are reported by element number.

A selectivity slot is named from the fleet's Selectivity: slot 2 of sel_inf is a descending inflection for the double-logistic family, the right-tail floor under DoubleNormal, and the age-1 selectivity under LogisticPM.

Under estimateMode = "Estimate" with any HCR but "NoFishing", object$obj is the projection object, whose only free parameters are log_Ftarget / log_Flimit, so the index describes those, not the hindcast. fit$convergence reads the hindcast index that fit_mod() captured before the projection remapped the model, and checks it against the vector it is labelling either way.

See also

parameter_dictionary() for what each block means.