Nonparametric Statistics with Applications to Science and Engineering with R, 2nd Edition

15Bootstrap

Confine! I'll confine myself no finer than I am:

these clothes are good enough to drink in; and so be these boots too:

an they be not, let them hang themselves in their own straps.

William Shakespeare (Twelfth Night, Act 1, Scene III)

15.1 Bootstrap Sampling

The idea of resampling is relatively recent. There is little that seems honest or intuitive about simulating or resampling observations from our original data set as a way of gaining improved information about the population, in general. However, despite the apparent controversy, resampling methods like the bootstrap, jackknife, or even permutation tests are powerful tools for modern statistical inference now that we can handle the computational costs they bear.

Bootstrapping might be the most frequently used sort of resampling, and to some, it is the most controversial. Resampling means we take a random sample from the sample, as if your sampled data represented a finite population of size . This new sample (typically of the same size ) is taken by “sampling with replacement,” so some of the items from the original ...