Skip to contents

Bootstrap

Source: R/boot.R, R/boot.default.R, R/boot.formula.R
boot.Rd

Bootstrap a single variable or a grouped variable

Usage

boot(x, ...)

# Default S3 method
boot(
  x,
  group = NULL,
  statistic = mean,
  success = NULL,
  conf.level = 0.95,
  B = 10000,
  plot.hist = TRUE,
  plot.qq = FALSE,
  x.name = deparse(substitute(x)),
  xlab = NULL,
  ylab = NULL,
  title = NULL,
  seed = NULL,
  ...
)

# S3 method for class 'formula'
boot(formula, data, subset, ...)

Arguments

x

a numeric, logical, factor, or character vector. Logical, factor, and character vectors with exactly two unique values are converted to 0/1, and mean is used to compute the proportion.

...

further arguments to be passed to or from methods.

group

an optional grouping variable (vector), usually a factor variable. If it is a binary numeric variable, it will be coerced to a factor.

statistic

function that computes the statistic of interest. Default is the mean.

success

a character string naming the level of x to code as 1 when x is a logical, factor, or character variable. Defaults to NULL, which uses the second factor level (alphabetically) or TRUE for logical vectors.

conf.level

confidence level for the bootstrap percentile interval. Default is 95%.

B

number of times to resample (positive integer greater than 2).

plot.hist

logical value. If TRUE, plot the histogram of the bootstrap distribution.

plot.qq

Logical value. If TRUE, create a normal quantile-quantile plot of the bootstrap distribution.

x.name

Label for variable name

xlab

an optional character string for the x-axis label

ylab

an optional character string for the y-axis label

title

an optional character string giving the plot title

seed

optional argument to set.seed

formula

a formula y ~ g where y is a numeric vector and g a factor variable with two levels. If g is a binary numeric vector, it will be coerced to a factor variable. For a single numeric variable, formula may also be ~ y.

data

a data frame that contains the variables given in the formula.

subset

an optional expression indicating what observations to use.

Value

A vector with the resampled statistics is returned invisibly.

Details

Perform a bootstrap of a statistic applied to a single variable, or to the difference of the statistic computed on two samples (using the grouping variable). If x is a binary vector of 0's and 1's and the function is the mean, then the statistic of interest is the proportion.

Observations with missing values are removed.

Methods (by class)

  • boot(default): Bootstrap a single variable or a grouped variable

  • boot(formula): Bootstrap a single variable or a grouped variable

References

Tim Hesterberg's website https://www.timhesterberg.net/bootstrap-and-resampling

Author

Laura Chihara

Examples


#ToothGrowth data (supplied by R)
#bootstrap mean of a single numeric variable
boot(ToothGrowth$len)
#> 
#> 	** Bootstrap interval for mean 
#> 
#>  Observed  ToothGrowth$len : 18.81333 
#>  Mean of bootstrap distribution: 18.80062 
#>  Standard error of bootstrap distribution: 0.97955 
#> 
#>  Bootstrap percentile interval
#>     2.5%    97.5% 
#> 16.88325 20.73000 
#> 
#> 		*--------------*


#bootstrap difference in mean of tooth length for two groups.
boot(ToothGrowth$len, ToothGrowth$supp, B = 1000)
#> 
#> 	** Bootstrap interval for difference of mean 
#> 
#>  Observed difference of mean : OJ - VC = 3.7 
#>  Mean of bootstrap distribution: 3.63724 
#>  Standard error of bootstrap distribution: 1.86715 
#> 
#>  Bootstrap percentile interval
#>     2.5%    97.5% 
#> 0.068500 7.346833 
#> 
#> 		*--------------*


#same as above using formula syntax
boot(len ~ supp, data = ToothGrowth, B = 1000)
#> 
#> 	** Bootstrap interval for difference of mean 
#> 
#>  Observed difference of mean : OJ - VC = 3.7 
#>  Mean of bootstrap distribution: 3.6779 
#>  Standard error of bootstrap distribution: 1.91449 
#> 
#>  Bootstrap percentile interval
#>      2.5%     97.5% 
#> -0.117250  7.344333 
#> 
#> 		*--------------*


# Penguin Survival proportion
boot(penguin_survival$Status, data = penguin_survival, B = 1000)
#> Note: "Survived" coded as the success (1).
#> 
#> 	** Bootstrap interval for proportion 
#> 
#>  Observed  penguin_survival$Status : 0.28371 
#>  Mean of bootstrap distribution: 0.28307 
#>  Standard error of bootstrap distribution: 0.02398 
#> 
#>  Bootstrap percentile interval
#>      2.5%     97.5% 
#> 0.2359551 0.3286517 
#> 
#> 		*--------------*


# same as above, but with the formula syntax
boot(~Status, data = penguin_survival, B = 1000)
#> Note: "Survived" coded as the success (1).
#> 
#> 	** Bootstrap interval for proportion 
#> 
#>  Observed  Status : 0.28371 
#>  Mean of bootstrap distribution: 0.28373 
#>  Standard error of bootstrap distribution: 0.02449 
#> 
#>  Bootstrap percentile interval
#>      2.5%     97.5% 
#> 0.2387640 0.3287219 
#> 
#> 		*--------------*


# Penguin Survival if tagged vs. untagged
# bootstrap difference in proportions of survival
boot(penguin_survival$Status, penguin_survival$TagType, B = 1000)
#> Note: "Survived" coded as the success (1).
#> 
#> 	** Bootstrap interval for difference of proportion 
#> 
#>  Observed difference of proportion : Electronic - Metal = 0.16218 
#>  Mean of bootstrap distribution: 0.15984 
#>  Standard error of bootstrap distribution: 0.04596 
#> 
#>  Bootstrap percentile interval
#>       2.5%      97.5% 
#> 0.06869911 0.24754380 
#> 
#> 		*--------------*


# now using the formula syntax
boot(Status ~ TagType, data = penguin_survival, B = 1000)
#> Note: "Survived" coded as the success (1).
#> 
#> 	** Bootstrap interval for difference of proportion 
#> 
#>  Observed difference of proportion : Electronic - Metal = 0.16218 
#>  Mean of bootstrap distribution: 0.16076 
#>  Standard error of bootstrap distribution: 0.04636 
#> 
#>  Bootstrap percentile interval
#>       2.5%      97.5% 
#> 0.07053987 0.25033504 
#> 
#> 		*--------------*