Skip to contents

Bootstrap tvcQGcomp Fits

bootstrap_tvcqgcomp.Rd

Run a subject-level cluster bootstrap for tvcQGcomp. In regular mode the function returns replicate-level meta-model coefficients. In audit mode it also stores the per-replicate bootstrap data, Monte Carlo data, natural course output, intervention outputs, and risk curves. The canonical user-facing survival bootstrap entry point is tvcQGComp_survival_boot().

Usage

tvcQGComp_survival_boot(
  data,
  config,
  n_boot = 200L,
  mc_size = 10000L,
  intervention_scenarios = NULL,
  natural_course = config$natural_course,
  seed = 1234L,
  replace_mc = NULL,
  meta_formula = NULL,
  meta_family = NULL,
  meta_target = config$meta_target,
  fit_meta = NULL,
  fit_exposure_models = NULL,
  exposure_scale = config$exposure_scale,
  quantization_breaks = NULL,
  joint_effect_fn = NULL,
  parallel = FALSE,
  n_workers = 2L,
  batch_size = NULL,
  checkpoint_file = NULL,
  resume_from_checkpoint = TRUE,
  verbose = TRUE,
  stop_on_error = TRUE,
  audit_mode = FALSE,
  keep_boot_objects = FALSE,
  keep_boot_curves = FALSE
)

summarize_tvcqgcomp_bootstrap(boot_res)

Arguments

data, config, mc_size, intervention_scenarios, natural_course, seed, replace_mc, meta_formula, meta_family, meta_target, fit_meta, fit_exposure_models, exposure_scale, quantization_breaks, joint_effect_fn

See tvcQGComp_survival(). When multiple meta targets are requested, bootstrap result columns are prefixed by target, for example coef_HR_joint_effect, coef_HD_joint_effect, and coef_RR_joint_effect. When fit_meta = FALSE, replicate outputs contain scenario-specific final risks instead of meta-model coefficients.

n_boot

Number of bootstrap replicates.

parallel

Logical; whether to evaluate bootstrap replicates in parallel.

n_workers

Number of workers when parallel = TRUE.

batch_size

Number of replicates per parallel batch.

checkpoint_file

Optional RDS file to update during the bootstrap run.

resume_from_checkpoint

Logical; if TRUE and checkpoint_file exists, resume from the last completed bootstrap replicate stored in the checkpoint.

verbose

Logical; print bootstrap progress.

stop_on_error

Logical; if TRUE, stop the bootstrap at the first failed replicate after saving a checkpoint. If FALSE, failed replicates are retained with an error: ... status and the bootstrap continues.

audit_mode

Logical; if TRUE, keep replicate objects for auditing.

keep_boot_objects

Logical; explicit switch to retain audit objects.

keep_boot_curves

Logical; retain a lightweight per-replicate curve table containing time-specific risk and survival without storing the full bootstrap audit objects.

boot_res

Result from tvcQGComp_survival_boot().

Value

tvcQGComp_survival_boot() returns either a data table of replicate meta-model coefficients or scenario-specific final risks or, when curve summaries and/or audit objects are retained, a list with results and optional curve_results, audit_objects, and audit_mode. summarize_tvcqgcomp_bootstrap() returns a summary table with means, standard errors, and percentile intervals.