Statistics Glossary → Bootstrap (resampling)
Bootstrap (resampling)concept
A method that draws thousands of new samples with replacement from the sample at hand to compute the uncertainty of a statistic directly.
Many confidence intervals rely on a formula and an assumption about the distribution of the statistic (e.g. normality). The bootstrap uses computing power instead: many new samples of the same size are drawn from the observations with replacement (an observation can be chosen more than once), and the statistic is recomputed in each.
The distribution of the thousands of values obtained estimates the sampling distribution of the statistic. In the percentile method the 95% confidence interval is the 2.5th and 97.5th percentiles of these values. Because the result depends on random draws, the number of replicates (B) is chosen large enough and a fixed seed makes the results reproducible.
An everyday example
Imagine a bag with cards showing the heights of 50 people. You draw a card, write it down, put it back, and repeat this 50 times; you note the mean of this new set of '50 people'. Doing this thousands of times shows how much the mean fluctuates, giving a confidence interval for the mean without any formula.
In MerQur
The 'Bootstrap CI' option of the t-tests (B = 5000, fixed seed), the Bootstrap Confidence Interval analysis, and bootstrap standard error options in many models.