Dimensionality Reduction for Countermovement Jump Metrics
Lachlan P. James, Haresh Suppiah, Michael R. McGuigan, David L. Carey
Abstract
PURPOSE: Dozens of variables can be derived from the countermovement jump (CMJ). However, this does not guarantee an increase in useful information because many of the variables are highly correlated. Furthermore, practitioners should seek to find the simplest solution to performance testing and reporting challenges. The purpose of this investigation was to show how to apply dimensionality reduction to CMJ data with a view to offer practitioners solutions to aid applications in high-performance settings. METHODS: The data were collected from 3 cohorts using 3 different devices. Dimensionality reduction was undertaken on the extracted variables by way of principal component analysis and maximum likelihood factor analysis. RESULTS: Over 90% of the variance in each CMJ data set could be explained in 3 or 4 principal components. Similarly, 2 to 3 factors could successfully explain the CMJ. CONCLUSIONS: The application of dimensional reduction through principal component analysis and factor analysis allowed for the identification of key variables that strongly contributed to distinct aspects of jump performance. Practitioners and scientists can consider the information derived from these procedures in several ways to streamline the transfer of CMJ test information.