6 141 GENERAL DISCUSSION There are multiple reasons why datasets may be incomplete, such as the absence of confounding variables or incomplete observations, resulting in data gaps that can have a negative impact on predictive models. This issue is broader than the missing variables in Chapters 5A and 5B. For example, in Chapter 2, additional information was desired to include variables that may have influenced a person’s daily walking routine, such as work schedule, days off, or meetings during lunch breaks. Similarly, in Chapter 3, for example, the ranking or score of a football match was unknown, as were any system changes, which reduced the informative value of the prediction. STRENGTHS AND LIMITATIONS, AND RECOMMENDATIONS The research presented in this thesis possesses several strengths and limitations. A major strength of our research was the use of existing large, personalised datasets of physical activity and physical performance. Large datasets can help to identify patterns and trends in physical activity and physical performance data that may be difficult to detect in smaller samples. In addition, utilising datasets comprising information on individual participants has facilitated the ability to generate detailed insights and predictions at the individual level, demonstrating the feasibility of more informed intervention on an individual level. Furthermore, another strength is the implemented various machine learning algorithms and advanced statistical techniques, including Targeted Maximum Likelihood Estimation allowing for identifying complex relationships and patterns that may be difficult or impossible to discern using traditional statistical techniques. This innovative approach has not been previously implemented in the field of sport science, and the results of this study demonstrate that integrating these techniques expands the methodological toolbox available to researchers in sport science. Furthermore, the various machine learning algorithms allow for the development of more accurate and reliable models for predicting physical performance in individuals. An additional strength is that it marks the first application of both a causal roadmap and a causal model in the field of sport science. The applied causal roadmap and causal model substantially advance understanding and analysing complex phenomena like in elite sports. By incorporating the principles of causality, using the causal roadmap and a causal model allows for a more thorough examination of the underlying mechanisms and factors that influence athletic performance. Overall, this approach holds promise for advancing our understanding and developing more rigorous and informative research.
RkJQdWJsaXNoZXIy MjY0ODMw