Practitioners in today's sport science scene are often becoming awash in high-dimensional data sets. From an athlete monitoring perspective, teams may exploit several mediums concurrently; from force platforms and hydration to wellness, GPS and more - each contributing numerous, if not hundreds, of variables. How are practitioners supposed to select markers that meaningfully influence their objective? The purpose of this document is to showcase a realistic methodology for extracting predictive variables from a soup full of options by employing elementary machine learning principles. Practical Applications . Pearson's correlation only detects linear relations, which is a major limitation when describing non-linear biological systems. Alternatively, distance correlation is sensitive to both linear and non-linear dependencies. . Information-theoretic measure, entropy, provides an inductive approach for sifting out predictive bivariate relationships from noisy, high-dimensional data sets.
© Copyright 2018 Sport Performance & Science Reports. Sport Performance & Science Reports. All rights reserved.