percboot: Perform Percentile Bootstrap for a Given Statistic

In ddarmon/clp: R Package for Computing Confidence Functions to Accompany CLP

Description

Approximate the bootstrap distribution for the percentile bootstrap, given a statistic as would be supplied to boot from the boot package.

Usage

percboot(
  data,
  statistic,
  B = 2000,
  sim = "ordinary",
  stratified = FALSE,
  ran.gen = function(d, p) d,
  mle = NULL,
  formula = NULL
)

Arguments

data	the dataframe or data matrix
statistic	a function that computes the statistic from data
B	the number of bootstrap replicates
sim	either "ordinary" (for case resampling bootstrap) or "parametric" (for parametric bootstrap)
stratified	whether or not (default) to use the use stratified sampling for the bootstrapping.
ran.gen	a function returning a random sample, for the parametric bootstrap
mle	the maximum likelihood estimate from the original sample, for the parametric bootstrap
formula	for use with interfacing to lm, glm, etc.

Value

A list containing the sample estimate, bootstrap estimates, and bootstrap CDF for the percentile bootstrap.

References

Thomas J. DiCiccio and Bradley Efron. "Bootstrap confidence intervals." Statistical Science (1996): 189-212.

Examples

# Bootstrap confidence functions for a single mean.

t.one.sample <- function(data, id = 1:length(data)){
  dat <- data[id]

  d <- mean(dat)

  return(d)
}

data(dietstudy)

bc <- percboot(data = dietstudy$weightchange[dietstudy$diet == 'Low Carb'],
        statistic = t.one.sample,
        B = 2000)

ddarmon/clp documentation built on Jan. 25, 2021, 6:22 p.m.