Making Bootstrap Statistical Inferences: A Tutorial

Abstract

Bootstrapping is a computer-intensive statistical technique in which extensive computational procedures are heavily dependent on modern high-speed digital computers. The payoff for such intensive computations is freedom from two major limiting factors that have dominated classical statistical theory since its beginning: the assumption that the data conform to a bell-shaped curve, and the need to focus on statistical measures whose theoretical properties can be analyzed mathematically. The name “bootstrap” was derived from an old saying about pulling oneself up by one's own bootstraps. In this case, bootstrapping means redrawing samples randomly from the original sample with replacement. The key idea, computations, advantages, limitations, and application potential of bootstrapping in the field of physical education and exercise science are introduced and illustrated using a set of national physical fitness testing data. Finally, an example of a bootstrapping application is provided. Through a step-by-step approach, the development and implementation of the bootstrap statistical inference are illustrated.