Degree Name

PhD (Doctor of Philosophy)

Program

Sport Physiology and Performance

Date of Award

5-2025

Committee Chair or Co-Chairs

Michael Ramsey

Committee Members

Hugo Espinosa, Kevin Carroll, Andrew Dotterweich

Abstract

This dissertation seeks to evaluate the reliability, validity, and utility of accelerometers and GNSS in sport-like scenarios, assess accelerometry-derived metrics and GNSS for distance prediction, and explore how accelerometer placement, specifically chest and back, affect the detection of various locomotor events. In Study I, the purpose was to investigate the accelerometer reliability and predictive capabilities of accelerometers and GNSS to predict a known distance. Thirty physically active female participants (mean height = 163.6 ± 6.8 cm, mean weight = 57.8 ± 8.7 kg) completed jogging trials over two courses (5m and 20m). Each participant wore two tri-axial accelerometers (100Hz) and a GNSS sensor (10Hz). Data were collected across total distances ranging from 20 to 165 meters in 5m increments. Inter-device reliability was assessed via intraclass correlation coefficients (ICCs), which ranged from 0.93 to 0.97. Four accelerometry-derived metrics and GNSS data were used to generate linear regression models to predict known distance. All four accelerometry based models performed well (R² = 0.96–0.98, RMSE = 5.72–6.17 m), while the GNSS model performed poorly (R2 = 0.306-0.671; RMSE = 36.1-24.8 m). In Study II, the purpose was to assess the influence of sensor placement (chest vs. back) on accelerometry-derived metrics and locomotor event detection. Thirty physically active female participants each wore two tri-axial accelerometers (100Hz), one at the xiphoid process (CHEST) and one between the shoulder blades (BACK). Each participant completed two 20m trials of walking, jogging, and running (6 total trials), plus 1–10 jumps and bounds. Across the 20m trials, good inter-device reliability was demonstrated for three accelerometry-based metrics Magnitude, Sum, and Impulse Load (CV = 0.83–9.83%), while Player Load was less reliable (CV = 6.32–17.00%). CHEST and BACK showed high positive predictive value (PPV > 90%) for step detection during all locomotor conditions but substantially lower PPV for jumping and bounding (

Document Type

Dissertation - unrestricted

Copyright

Copyright by the authors.

Download

Share

COinS