Glance at a FIMSFit object — one-row model summary

Returns a single-row tibble of model-level diagnostics following the generics::glance() convention. Standard information-criterion columns (logLik, AIC, BIC, nobs) are included alongside FIMS-specific diagnostics (max_gradient, marginal_nll, total_nll, converged, terminal_ssb).

Usage

# S3 method for class 'FIMSFit'
glance(x, ...)

Arguments

x: A FIMSFit object returned from fit_fims().
...: Unused; present for S3 method compatibility.

Value

A tibble::tibble() with one row and the following columns:

logLik: Marginal log-likelihood at the MLE (-opt$objective).
AIC: Akaike information criterion (marginal, fixed effects only).
BIC: Bayesian information criterion (marginal, fixed effects only).
nobs: Total number of data observations (rows with both an observed and an expected value in get_estimates()).
npar_fixed: Number of fixed-effect parameters.
npar_random: Number of random-effect parameters.
marginal_nll: Marginal negative log-likelihood (opt$objective); NA if the model was not optimized.
total_nll: Total (joint) negative log-likelihood from the TMB report (report$jnll).
max_gradient: Maximum absolute gradient at the MLE. Values below 0.001 are generally considered well-converged.
converged: Logical; TRUE when opt$convergence == 0 and max_gradient < 0.001.
terminal_ssb: Spawning stock biomass at the terminal year, extracted from the TMB report. Returns a list-column when multiple populations are present.
fims_version: The version of FIMS used to fit the model.
runtime_secs: Total wall-clock time of the fit in seconds.

Information criteria

AIC and BIC are computed from the marginal log-likelihood (i.e. after integrating out random effects), using the number of fixed-effect parameters as $k$: $$AIC = 2k - 2 \ell_m$$ $$BIC = k \log(n) - 2 \ell_m$$ where $\ell_m$ is the marginal log-likelihood and $n$ is the total number of data observations. These values will be NA when optimize = FALSE was passed to fit_fims().

Examples

if (FALSE) { # \dontrun{
data("data_big")
data_4_model <- FIMSFrame(data_big)

fit <- create_default_parameters(
  configurations = create_default_configurations(data = data_4_model),
  data = data_4_model
) |>
  initialize_fims(data = data_4_model) |>
  fit_fims(optimize = TRUE)

glance(fit)

# Compare sensitivity runs in one table
dplyr::bind_rows(
  glance(age_only_fit) |> dplyr::mutate(model = "age_only"),
  glance(length_only_fit) |> dplyr::mutate(model = "length_only")
)
} # }