Skip to contents

Resample data

Source: R/resample.R
resample.Rd

Function to allow various resampling of data, and application to a function as desired. Choices for simple, stratified, and block (both sliding and separate) resampling. Bootstrap and permutation resampling can be performed.

Usage

resample(
  data,
  M = 1000,
  resample_blocks = c("separate", "sliding"),
  replace = TRUE,
  blocksize = 1,
  strata = NULL,
  sizes = NULL,
  fn = NULL,
  ignore.columns = NULL,
  ...
)

Arguments

data

Data.frame (or vector) to be resampled where the rows are the observations and the columns are the variables. Note, all variables are permuted not specified in ignore.columns.

M

Numeric. Number of resample iterations.

resample_blocks

String indicating method of selecting resample blocks. Options are 'separate' and 'sliding'. When blocksize is 1, there is no difference. Sliding take sliding groups, thus repeating values, i.e. 1-4, 2-5, 3-6, and so on. Separate does not repeat the values, i.e. 1-4, 5-8, 9-12, and so on. Size of each block is defined using blocksize.

replace

Boolean. Indicates if the data should be permuted (FALSE) or bootstrapped (TRUE). Note that permutation may be impossible with large sizes.

blocksize

Numeric for the size of the blocks.

strata

String or numeric. This indicate the column to stratify the data when method is stratify. This can be the column number or the column name. When NULL the data is not stratified. When given, strata are sampled separately.

sizes

Option for selecting the resampled size or sizes of each stata. When used for non-strata, either numeric or NULL and is taken as the number of observations. When used for strate, can be numeric (single value or a value for each strata), function (e.g. min or max), or NULL (original sizes).

fn

Function to apply on the resampled data. When NULL, the resampled data are directly returned.

ignore.columns

Name or column numbers to ignore when resampling data. These are not permuted. Note that if less/more samples are collected than the original, these are permuted separately.

...

Additional parameters for fn.

Value

Resampled data sets or function applied to resampled data.

Examples

results <- resample(1:100)
results <- resample(1:100, fn = mean)

n <- 50
data <- data.frame(
  output = NA,
  predictor1 = rnorm(n),
  predictor2 = rnorm(n)
)
data$output <- 2 * data$predictor1 - data$predictor2 + stats::rnorm(nrow(data))
results <- resample(data, ignore.columns = "output")

fn <- function(data) {
  coef(lm(output ~ ., data = data))
}
# Warning, M is below recommendation for example speed
results <- resample(data = data, fn = fn, M = 10, ignore.columns = "output")