Bootstrap a fitted dose-response model — bootstrap_dose

Generates grouped binomial bootstrap samples and refits the model. The "observed" resampling method preserves the legacy CAMRA behavior by using each dose group's observed response probability. The "fitted" method is a model-based parametric bootstrap. Mirai runs show progress as batches finish.

Usage

bootstrap_dose_response(
  object,
  times = 10000L,
  resample = c("observed", "fitted"),
  seed = NULL,
  backend = c("auto", "sequential", "mirai"),
  compute = NULL,
  workers = NULL
)

Arguments

object: A qdr_fit object.
times: Number of bootstrap replicates.
resample: Either "observed" or "fitted".
seed: Optional integer random seed.
backend: Execution backend. "auto" uses mirai above 1,000 replicates when it is installed and otherwise runs sequentially. "sequential" runs refits in the current R process. "mirai" distributes refits to existing daemons or a temporary local pool. When loaded with pkgload::load_all(), the same checkout is loaded on each worker.
compute: Optional mirai compute profile name. Ignored by the sequential backend.
workers: Number of temporary local mirai daemons. NULL uses 75% of detected physical cores, with a minimum of one. Ignored when using existing daemons or the sequential backend.

Value

A qdr_bootstrap tibble with one row per replicate.