The resampling cross-validation technique in exercise science: modelling rowing power.

Abstract

The past 10-15 yr have witnessed a rapid increase in the development of new (and not so new) statistical methods that capitalize on recent advances in high-speed computing. These computer-intensive methods are often broadly referred to as resampling techniques and take several forms depending on the specific details of the procedure and the information of interest. Resampling techniques can be used both for inferential hypothesis testing as well as exploratory data description. Regardless of which method is employed, the central unifying theme is based upon the computer's power to rapidly resample many pseudosamples from a known (in-hand) data set (e.g., randomization tests, jackknife, boot-strap, cross-validation) or to randomly generate many pseudosamples from a theoretical probability distribution (e.g., normal, binomial, Poisson) with some known parameters (Monte Carlo method). This paper is not intended as a detailed description of computer-intensive methods, but only as an introduction to the resampling approach in cross-validation. A brief discussion of the motivation and an example in an exercise science context will be presented.