In Aehmlo/strop: Bootstrap Methods for Inference

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)
library(strop)

General bootstrap functionality

At the core of strop is the bootstrap function. This is a generic function with a default method (for one-dimensional data) and a method for data frames (two-dimensional data). This function is a constructor for an object of class strop, which can be used in the package's other provided functions.

Single-valued bootstrap

Single-valued bootstrapping is straightforward: the default function used by bootstrap is mean.

data(mtcars)
bootstrap(mtcars$hp)$pop

Multivalued bootstrap

Both the data frame and default constructors may be used with multivalued functions.

cen <- function(x) list(mean = mean(x), median = median(x))
bootstrap(mtcars$hp, cen)$pop

Calculating bootstrap quantiles

strop provides a quantile.strop function for calculating quantiles based on the bootstrap procedure. This function simply delegates to quantile(obj$stats).

quantile(bootstrap(mtcars$hp), c(0.05, 0.95))

Calculating bootstrap confidence intervals

strop exposes a confint.strop function for calculating central confidence intervals based on the bootstrap procedure.

confint(bootstrap(mtcars$hp, n = 100))

Printing bootstrap results

strop provides a print.strop function for assessing the results of bootstrap iteration. The function call, (original) sample estimates, and the beginning of the list of bootstrap estimates are printed.

print(bootstrap(mtcars$hp, n = 100))

Visualizing bootstrap results

strop provides plot.strop function to visualize the results of a bootstrap application. A histogram of each statistic is plotted for multi-valued bootstrapping, and central confidence intervals are indicated on each plot.

plot(bootstrap(mtcars$hp, n = 100))

Multiple linear regression

strop provides a convenience function for multiple linear regression analysis using the bootstrap procedure: mlrboot. Given a fitted model object, mlrboot produces a bootstrap object for the estimation of model coefficients.

fit <- lm(hp ~ ., data = mtcars)
boot <- mlrboot(fit)
print(boot, max = 5)
plot(boot, vars = c("cyl", "wt"))

Aehmlo/strop documentation built on Dec. 1, 2019, 2:55 a.m.