Skip to contents

Convert a grid of MCMC cells to a tmbfit object

Source: R/grid.R
grid_mcmc_to_tmbfit.Rd

Builds a balanced posterior sample from saved grid-cell MCMC files and stores the combined draws in a tmbfit object. Fixed grid parameters such as h, psi, m0, and m10 are inserted into the sample matrix when the corresponding log-parameter columns are requested.

Usage

grid_mcmc_to_tmbfit(
  grid_files,
  grid_values = NULL,
  target_sample_names = NULL,
  target_par_names = NULL,
  allow_mixed_run_signatures = FALSE,
  n_draws = 2000L,
  seed = 1L
)

Arguments

grid_files

Character vector of canonical grid-cell .sbt.rds files. Legacy grid files are rejected because they lack the state diagnostics and scientific signatures required for combination.

grid_values

Optional data.frame of grid-cell values used when saved files do not contain grid_cell.

target_sample_names

Character vector of sample columns to keep.

target_par_names

Character vector of parameter columns that must be complete in the combined posterior. The default uses target_sample_names excluding sampler diagnostics such as lp__.

allow_mixed_run_signatures

Logical. The default, FALSE, requires every cell to share one sampler-run signature. Set to TRUE only when the calling workflow has independently validated an explicitly approved mixed-source selection. Grid-input and model signatures always remain uniform.

n_draws

Integer number of balanced draws to take across grid cells.

seed

Integer seed used for draw allocation and within-cell sampling.

Value

A list with fit (a tmbfit object), draw_metadata, draw_plan, samples, and the verified source_signatures. Canonical cells must share grid-input and model signatures and must have matching retained-posterior payload checksums. Run signatures are common unless an approved caller explicitly enables a mixed-source selection.

Details

This converter revalidates canonical structure, source signatures, posterior payloads, and biological states. It does not itself enforce the MLE-gradient, estimability, R-hat, ESS, divergence, or tree-depth acceptance thresholds. The calling production workflow must accept every cell's diagnostic row before combination. Supplying seed changes the global random-number state.

See also

Other grid workflows: grid_to_tmbfit(), make_mle_grid_parameters(), run_grid(), run_grid_mcmc(), sample_grid(), select_balanced_grid_iters(), summarise_mle_grid(), validate_mle_grid_state_records()