Parametric bootstrap for a within-group SensIAT model

Usage

parametric_bootstrap_within_group(
  within_group_model,
  nboot = 100,
  simulate_args = list(),
  seed = NULL,
  progress = interactive(),
  sample_coefficients = FALSE,
  refit = TRUE,
  return = c("coefficients", "models", "data"),
  prune_models = FALSE,
  gc_every = 10L,
  verbosity = c("none", "basic", "detailed"),
  verbose = NULL
)

Arguments

within_group_model: A fitted SensIAT_within_group_model object.
nboot: Number of bootstrap replicates.
simulate_args: List of arguments to pass to simulate_SensIAT_data(). If not specified, End, n_subjects, initial_outcome_mean, and initial_outcome_sd are inferred from the fitted model.
seed: Optional seed for reproducibility.
progress: Logical; show progress bar when available.
sample_coefficients: Logical; if TRUE, sample coefficients from an asymptotic multivariate normal distribution when vcov() is available. If FALSE (default), use original fitted coefficients.
refit: Logical; if TRUE (default), fit a SensIAT_within_group_model on each simulated replicate using the original model's settings.
return: One of "coefficients" (default), "models", or "data". "coefficients" is memory-efficient and stores only replicated marginal mean coefficients.
prune_models: Logical; when return = "models", prune each replicated model before returning.
gc_every: Integer. Run gc(FALSE) every gc_every replications. Use NULL to disable explicit garbage collection.
verbosity: Logging verbosity for bootstrap internals: one of "none", "basic", or "detailed".
verbose: Deprecated shortcut; if TRUE, equivalent to verbosity = "detailed".

Value

If return = "coefficients", a SensIAT_withingroup_bootstrap_results object. Otherwise returns a list of replicated fitted models ("models") or simulated datasets ("data").

Examples

if (FALSE) { # \dontrun{
data("SensIAT_example_data", package = "SensIAT")

# Fit a single-index outcome model on a small subset of the example data.
small_data <- dplyr::filter(
  SensIAT_example_data,
  Subject_ID %in% head(unique(SensIAT_example_data$Subject_ID), 8)
)

model <- fit_SensIAT_within_group_model(
  group.data = small_data,
  outcome_modeler = fit_SensIAT_single_index_fixed_coef_model,
  alpha = 0,
  id = Subject_ID,
  outcome = Outcome,
  time = Time,
  End = 830,
  knots = c(60, 260, 460)
)

# This example may take a long time because it fits a single-index outcome model
# and generates bootstrap replicates.
res <- parametric_bootstrap_within_group(
  nboot = 2,
  within_group_model = model,
  simulate_args = list(
    n_subjects = 3,
    End = 5,
    max_visits = 5
  ),
  seed = 123,
  refit = TRUE
)

print(res[[1]])
} # }