Capture a linkage specification

Usage

linkage_spec(
  formula,
  param = NULL,
  data = NULL,
  by = ~species,
  species = NULL,
  sex = NULL,
  link = "log",
  init = NULL,
  bounds = NULL,
  priors = NULL,
  re_group = NA_character_,
  est_phase = 1L
)

Arguments

formula: one-sided R formula whose RHS describes the linear predictor for param (e.g. ~ 1, ~ temp, ~ temp + PDO).
param: target parameter name on the natural scale (e.g. "alpha", "M1", "K"). May be NULL when the spec is built inside a build_*() call that infers the parameter name from the enclosing list key (see build_growth()).
data: (Optional) data frame for formula validation. Currently validation happens at materialization time inside fit_mod().
by: one-sided formula naming stratifying factors that should each get their own coefficients. Allowed names are species, sex, and age_bin. The default ~species produces one coefficient set per species (the typical multispecies assessment use case); pass ~species + sex for per-(species, sex) coefficients, or NULL to share a single coefficient set across every species/sex.
species: optional integer vector of 1-based species ids that this spec applies to. NULL (default) means every species in strata$species at materialization time. Use this to give different species different formulas, e.g. by registering multiple specs against the same parameter – see build_growth() for the multi-spec syntax.
sex: optional vector of sex ids that this spec applies to. May be supplied as integers (1L = female, 2L = male) or as character strings ("Females"/"Males", case-insensitive; "female", "male", "f", "m" are also accepted). NULL (default) means every sex in strata$sex at materialization time. Only meaningful when by includes sex; otherwise the filter is a no-op. Use this to register separate specs per sex (e.g. one prior on females, another on males) against the same parameter.
link: link function relating the linear predictor to the natural-scale target parameter. One of "log" (default) or "identity". With link = "log", log(param) = X * beta – slope contributions are multiplicative on the natural-scale parameter. With link = "identity", param = X * beta – slope contributions are additive on the natural scale. All linkage targets currently expose log-scale TMB parameters, so "log" is the natural default; "identity" is reserved for future processes (e.g. logit for steepness).
init: optional named numeric vector of initial values keyed by the design-matrix column name (e.g. c((Intercept) = 4, temp = 0)). Missing entries default to 0.
bounds: optional named list of c(lower, upper) keyed the same way as init.
priors: optional named list whose entries are Rceattle_priors objects, keyed by design-matrix column name. Inside this argument the unprefixed shorthand normal(), lognormal(), gamma(), and beta() resolves to the corresponding prior_* constructors via a private data mask, so priors = list(temp = normal(0, 1)) works without masking base::gamma() or base::beta() at the package level. Equivalent to priors = list(temp = prior_normal(0, 1)).
re_group: optional character: name of a random-effect grouping for these coefficients. NA (default) means fixed.
est_phase: optional integer estimation phase. Default 1L.

Value

An Rceattle_linkage_spec object.