Issues in the application of statistical techniques in sport and exercise science.

The aim of this research is to demonstrate the benefits and limitations of selected techniques used to analyse data derived from , sport and exercise science research. Although statistical techniques are easy to access through software packages, supporting literature about their appropriate application is less common. Many researchers are unaware of the full benefits or potential pitfalls when using these techniques. An understanding of the appropriate use of statistics will benefit the researcher by maximising the potential for analysing, interpreting and applying data correctly. Furthermore, it will minimise wasted effort or dissemination of inaccurate information through incorrect analyses. In this thesis examples are derived from fifteen published articles based on five topics that illustrate the appropriate use of particular statistical techniques. Firstly, the use of 'agreement' and 'least-products-regression' as appropriate techniques for comparing repeated measures are demonstrated (e.g. Mullineaux et al., 1999). Both techniques revealed that over two separate days the peak-torque-extension of the knee of healthy females is unreliable. Secondly, the use of 'allometric' scaling of body size differences that should allow for meaningful comparisons between participants' measurements is explored (e.g. Batterham, George and Mullineaux, 1997). Results showed that left ventricular mass is related to fat free mass to the power of 1.07 (0.92 to 1.22; 95% Cl). Thirdly, mathematical modelling is used to explore a theory that would be difficult to test empirically (e.g. Payton, Hay and Mullineaux, 1997). Results revealed that body roll contributes substantially to the propulsive force in front crawl swimming. Fourthly, logistic regression is used to predict group membership from the combined effect of several independent variables (e.g. Mullineaux et al., 2001a). It was found that the likelihood of participation in adequate physical activity to promote health can be strongly predicted from six variables. Lastly, in an invited review paper, key features in the application of research methods and statistics in biomechanics and motor control are highlighted (e.g. Mullineaux et al., 2001b). These published papers form a body of work that will facilitate a greater and more appropriate use of selected statistical techniques in sport and exercise science.