bootstrap: bootstrap

Description

Generate a vector of bootstrap samples.

Usage

bootstrap(
  x,
  trt = NULL,
  trt_control = "Control",
  FUN = NULL,
  varname = NULL,
  varcol = NULL,
  arglist = NULL,
  n_samples = 1
)

Value

If FUN is NULL returns a vector of objects of class Bootstrap. If FUN is non-NULL returns a vector of objects of class BootstrapStatistic

Arguments

x: Source data to bootstrap.
trt: Treatment variable. (optional)
trt_control: Value for treatment control arm. Default value is 'Control'.
FUN: Function to compute statistic for each bootstrap sample. (optional)
varname: Name of variable in x on which to compute FUN. If x has only one column varname is not needed. If x has more than one column then either varname or varcol must be specified.
varcol: Column index of x on which to compute FUN. If x has only one column varcol is not needed. If x has more than one column then either varname or varcol must be specified.
arglist: List of additional arguments to pass to FUN.
n_samples: Number of bootstrap samples to generate.

Details

Each bootstrap sample will retain the in-bag and out-of-bag data. Optionally, the user may specify a function to compute a statistic for each in-bag and out-of-bag sample. This function may be a built-in R function (e.g. mean, median, etc.) or a user-defined function (see Examples). If no statistic function is provided bootstrap returns a vector of objects of class Bootstrap. If a statistic function is provided bootstrap returns a vector of objects of class BootstrapStatistic, which in addition to the in-bag and out-of-bag samples contains the name of the statistic, variable on which the statistic is computed, and the numerical result of the statistic for each in-bag and out-of-bag sample.

Examples

Run this code

## Generate example data frame containing response and treatment
N <- 20
x <- data.frame( runif( N ) )
names( x ) <- "response"
x$treatment <- factor( sample( c("Control","Experimental"), size = N,
                       prob = c(0.8,0.2), replace = TRUE ) )

## Generate two bootstrap samples without regard to treatment
ex1 <- bootstrap( x, n_samples = 2 )

## Generate two bootstrap samples stratified by treatment
ex2 <- bootstrap( x, trt = x$treatment, trt_control = "Control", n_samples = 2 )

## For each bootstrap sample compute a statistic on the in-bag and out-of-bag data
ex3 <- bootstrap( x, FUN = mean, varname = "response", n_samples = 2 )

## Specify a user-defined function that takes a numeric vector input and
## returns a numeric result
sort_and_rank <- function( z, rank ){
  z <- sort( z )
  return( z[rank] )
}

ex4 <- bootstrap( x, FUN = sort_and_rank, arglist = list( rank = 1 ),
                  varname = "response", n_samples = 2 )