Bootstrap and Jackknife, Overview

A quarter of a century ago, the advent of fast, cheap computing saw the emergence of computationally intensive methods that sought to exploit the data itself as a resource for information rather than relying on restrictive classical distributional assumptions that often failed to hold in practice. Methods such as the whimsically named jackknife and bootstrap have found wide use among practitioners in a variety of contexts. In this article, the basic uses of resampling techniques are described, with particular attention to bias and variance estimation and confidence interval construction. Implementation of these techniques in popular statistical software is also described.