An Introduction to the Bootstrap Method

Ms. Gitte Kremling
Graduate Teaching Assistant
University of Wisconsin – Milwaukee

“Assume we are interested in the average length of fish in Lake Michigan. Since we cannot measure the length of all fish, we only sample a tiny part of the total population and compute their mean length. In order to reason about the average length of the whole population, e.g. to construct a confidence interval for it, we need to get a feeling for the variability of the mean that we have computed. For that purpose, the classical bootstrap method creates many new samples by resampling from the original data set of heights and then computes the mean for all of these new samples. Given this data, we can create a histogram of bootstrap means which provides an estimate for the distribution of the sample mean. In this talk, we will motivate and explore the classical, parametric and residual bootstrap methods.”

Click to view flyer.