Dr Jonathan Henderson is a Research Fellow at Queen’s University Belfast within the Mathematical Sciences Research Centre. His research focuses on the development and application of robust statistical methods, with recent work exploring mixed-effects models in sports science. He completed his PhD at Queen’s University Belfast, where his research centred on spatial statistics and the modelling of retinal cell loss in glaucoma.
Recent years have seen a rise in data analysis within sports science, particularly in evaluating player performance and factors affecting team success. The increased collection of data aims to provide teams with a competitive edge. Football’s growing popularity and financial influence have led to more matches, pushing players to their limits and increasing injury risks, including cardiovascular events and fatigue-related issues.
This study examines heart rate data from players of a semi-professional football team, offering insights into exertion and recovery during games. By analysing this data, we aim to detect abnormal physiological responses, such as signs of fatigue or heightened injury risk. These irregularities may indicate cardiovascular strain, muscle overuse, or other conditions that impact performance.
A linear mixed-effects model was initially applied to explore these variations but revealed violations of normality assumptions in residuals and random effects. To address this, we adopted robust mixed-effects models, which handle extreme data points more effectively and improve the accuracy of our findings. This approach ensures a more reliable analysis of real-world data.
We compare different robust models to determine their ability to identify players with unusual heart rate patterns, incorporating time-varying degrees of freedom in the t-distribution. Our focus is on evaluating their fit and predictive performance relative to conventional models.
By leveraging robust mixed-effects models, this study provides practical insights into player health and performance. Coaches, analysts, and medical staff can use these findings to identify players at risk of overexertion and implement targeted interventions, such as adjusted training or rest periods.
Overall, this research highlights the importance of advanced statistical methods in sports science. By applying robust models to heart rate data, we improve understanding of football’s physical demands, ultimately enhancing player safety, performance management, and long-term well-being.
Kevin Wilson is a Professor of Applied Statistics at Newcastle University. His research interests are in Bayesian analysis, particularly prior elicitation, uncertainty quantification and design of experiments. His main application areas are in engineering, sports and biostatistics, particularly diagnostics and cluster randomised trials. He have published in leading journals in statistics including Bayesian Analysis, the Journal of the Royal Statistical Society (RSS), Statistics in Medicine and Technometrics. He has conducted consultancy projects with Costello Medical, NICE, Rocket Medical, Northern Gas Networks, Mologic and the Department for Culture, Media and Sport, amongst others. He is a fellow of the Royal Statistical Society. He is principal investigator on the Leverhulme grant: Loss-based Bayesian additive regression trees and the EPSRC grant SaFEGen: A statistical framework for efficient evidence generation in diagnostic development. He enjoys running, and competes in road and cross country races for Wallsend Harriers.
A marathon is a race of 42.195km or 26 miles and 385 yards. The most prestigious marathons include the Olympic Games marathons (every 4 years) and the marathons at the World Athletics Championships (every 2 years). Each year there is a series of 6 marathon majors (soon to be 7) in Berlin, Boston, Chicago, London, New York and Tokyo. While the Olympics and World Championships are in different cities each time, the majors are run on the same course each year.
Marathons are different to shorter races - professional athletes can only run 2-3 a year. So not all of the best runners compete in the same event. Unlike track events the courses are very different - Boston is a hilly point-to-point route, whereas Berlin is a flat looped course. So it isn’t obvious which are the best marathon performances each year, and certainly not of all time.
This talk attempts to answer this question, for the men's and women's marathons, using linear mixed effects models fitted using Bayesian inference. The data consist of the winning times of each marathon major up to the end of 2024.
Is Eliud Kipchoge the greatest of all time (GOAT) in the men's marathon? Have super-shoes with carbon plates had a discernible effect on winning marathon times? Would Paula Radcliffe beat the top Kenyan women today? All will be revealed (maybe) in this talk.
Dr Ruth Salway has a background in statistical methodology and is currently an applied statistician in the Centre for Public Health at the University of Bristol. She works across a range of applied areas in public health, including physical activity and active travel, and specialises in the design and evaluation of complex public health interventions, especially in understanding population-level impacts over time. She is particularly interested in bridging the gap between theoretical methods and their use in practice, and has applied a range of novel statistical techniques to provide robust evidence that addresses complex questions which are essential for informing public health policy and improving health, especially at population level.
While cluster randomised controlled trials are common in evaluations of school-based physical activity interventions, stepped wedge designs, where schools are randomised to a sequence of measurements and each sequence transitions to intervention status at a different time, are not currently used. We explore the feasibility and statistical power of this design, by balancing practical and statistical considerations. This will inform the design of a stepped wedge evaluation for a school-based intervention to increase children's physical activity.
We first conducted a qualitative study to identify school constraints and included other practical considerations such as class sizes, which limit the cluster sizes, and seasonal patterns in children’s physical activity, which are important as stepped wedge designs are partially confounded by time. We then explored statistical power, via simulation in R, for a range of different scenarios and stepped wedge designs informed by these considerations.
Schools reported the need for an intervention implementation period and a maximum of 4 measurements per school year, which led to consideration of incomplete stepped wedge designs, where not all schools are measured at each point. However, seasonal patterns caused bias in some incomplete designs. Statistically, the best designs to reduce bias and increase power had a mix of control and intervention measurements in the middle of the study and a spread of measurements across the whole study duration. Power depended on a combination of the overall recruitment and retention rates.
Stepped wedge trials are a viable design for evaluating school-based physical activity interventions, requiring half the number schools and fewer overall measurements than a similarly powered cluster randomised controlled trial. Incomplete designs offer the flexibility to work around practical constraints, although care must be taken when seasonality is present. Findings will be relevant to the design of a wide range of school-based stepped wedge trials.