Create a staged, portable SBT model fit

With only processed data, or preprocessing inputs supplied through data_in, creates the first stage of the standard workflow: sbt_fit(data), sbt_optimise(), then sbt_mcmc(). data_in is passed to get_data() for the current model or get_data_v1() for the frozen V1 model. Missing parameters, priors, map, and bounds are resolved and stored when the objective is first built. The existing completed-fit form remains supported by supplying both obj and opt; that transient objective is used to capture the fitted state but is never serialized.

Usage

sbt_fit(
  data = NULL,
  obj = NULL,
  bounds = NULL,
  control = NULL,
  opt = NULL,
  estimability = NULL,
  diagnostics = list(),
  metadata = list(),
  mcmc = NULL,
  mcmc_settings = list(),
  keep_mcmc_raw = FALSE,
  full_mcmc_mle = FALSE,
  makeadfun_args = list(),
  optimizer = "nlminb",
  model = c("sbt_model", "sbt_model_v1"),
  data_in = NULL
)

Arguments

data

Processed model data returned by get_data() and used to construct obj. Supply exactly one of data and data_in.

obj

Fitted RTMB objective. This is a transient input and is never serialized.

bounds

Optional parameter bounds used for optimization.

control

Optional optimizer control list.

opt

Optimizer result, normally returned by stats::nlminb().

estimability

Optional result from check_estimability(). Only a compact summary is retained.

diagnostics

Optional named list of additional fit diagnostics.

metadata

Optional named list of user metadata.

mcmc

Optional SparseNUTS-style fit, posterior sample array, or matrix. Only portable sample and diagnostic fields are retained.

mcmc_settings

Optional named list recording sanitized sampler settings.

keep_mcmc_raw

If TRUE, also retain the original MCMC object when it is portable. The normalized sbt_mcmc fields remain canonical.

full_mcmc_mle

If TRUE, retain every element of the sampler's MLE payload. The default keeps only estimates, standard errors, and correlations.

makeadfun_args

Optional named list of additional arguments needed to rebuild RTMB::MakeADFun(). Reserved core arguments are not allowed.

optimizer

Character name of the optimizer used.

model

Model function identifier: "sbt_model" or "sbt_model_v1".

data_in

Optional preprocessing input list passed to get_data() or get_data_v1(), according to model.

Value

An object of class sbt_fit containing portable state only. Principal components are data, parameters, map, bounds, and configuration for the scientific setup; fit and mcmc for fitted results; posterior_summaries and msy for derived outputs; and validation, status, and provenance for acceptance and reproducibility metadata.

See also

Other fit workflow: check_mcmc(), check_mle(), sbt_add_bounds(), sbt_add_map(), sbt_add_msy(), sbt_add_parameters(), sbt_add_priors(), sbt_build_object(), sbt_example_fit(), sbt_fit_compatibility(), sbt_fit_io, sbt_fit_rebuild(), sbt_fit_report(), sbt_fit_validate(), sbt_fit_validation(), sbt_mcmc(), sbt_obj(), sbt_optimise()

Examples

data <- sbt_example_data()
fit <- sbt_fit(data, metadata = list(run = "documentation example"))
fit
#> <sbt_fit>
#>   Model:       sbt_model (schema 2)
#>   sbt version: 3.0.0
#>   Stage:       data
#>   Estimability: not run
#>   MCMC:        not stored
#>   MSY:         0 stored calculations
#>   Next:        sbt_optimise() or configuration setters
#>   Created:     2026-08-26 16:26:33 UTC
fit$configuration$origins
#> $parameters
#> [1] "unresolved"
#> 
#> $priors
#> [1] "supplied"
#> 
#> $map
#> [1] "unresolved"
#> 
#> $bounds
#> [1] "unresolved"
#> 
#> $random
#> [1] "default"
#> 
#> $makeadfun_args
#> [1] "default"
#>