Bootstrap IPCW weighted data for competing risks survival analysis

Draws B bootstrap samples from the original wide-format competing risks data, converts each sample to long (counting-process) format via wide_to_long_cr(), and appends IPCW weights via add_ipcw_weights_cr(). Returns the results as a list of long-format data frames that can be used directly with cuminc_ipcw_cr(), or further processed with fg_split_cr() and add_fg_weights_cr() for a Fine-Gray bootstrap.

Usage

get_ipcw_boot_cr(
  data,
  B = 500,
  time_var = "t",
  event_var = "delta",
  covariate = "z1",
  cens_level = "censor",
  event2_level = "event_2",
  strat = "no",
  seed = NULL
)

Arguments

data

A wide-format competing risks data frame containing the columns specified by time_var, event_var, and covariate.

B

Integer. Number of bootstrap samples. Default is 500.

time_var

Character string. Name of the event/censoring time column. Default is "t".

event_var

Character string. Name of the event indicator column. Default is "delta".

covariate

Character string. Name of the covariate column. Default is "z1".

cens_level

Character string. Factor level in event_var representing censoring. Default is "censor".

event2_level

Character string. Factor level in event_var representing the competing event (event type 2). Default is "event_2".

strat

Character. Passed to add_ipcw_weights_cr(). "no" (default) fits a single Cox model for the censoring distribution; "yes" estimates the censoring distribution non-parametrically within each stratum of covariate.

seed

Optional integer seed for reproducibility. Default is NULL.

Value

A list of B data frames, each in long (counting-process) format with an IPCW weight column p_notcens, as returned by add_ipcw_weights_cr().

Examples

if (FALSE) { # \dontrun{
set.seed(42)
dat <- sim_data_cr(n = 200, censoring = "baseline")
boot_list <- get_ipcw_boot_cr(dat, B = 50)
} # }