Research methods: Sample size and variability effects on statistical power

This chapter highlights factors affecting statistical power, and exploring methods to analyse time-series data in biomechanics. Topics related to statistical power include estimating sample size, reducing variability through ratio- and offset-normalizations, and combining multiple trials. Instead of reducing variability for statistical purposes, methods are covered to analyse the functional nature of variability, which in itself influences the control or technique and outcome or performance of movement. Methods include: confidence intervals for discrete univariate data, and CI2 for bivariate data; vector coding, cross-correlations and continuous-relative-phase for pairwise comparisons; and principal component analysis, functional principal-component analysis and statistical parametric mapping for more complex situations with multiple degrees of freedom. This chapter highlights factors affecting power, both through the research design and implementation of data transformations, whilst also focusing on methods to explore and analyse time-series data. It also focuses on a small area of research methods, many factors were highlighted that can affect results and consequently the meaning of data. Variability can be reduced through the research design and methods by, for example, using a more reliable measure, averaging several trials, removing noise through filtering or correcting for the violation of statistical assumptions. The sample statistics should therefore provide a reliable and unbiased predictor of the population parameter. Ratio normalisation is underpinned by the assumption that a theoretical and statistical relationship exists between the dependent variable and a covariate. A phase-plane portrait can be used to calculate variability using continuous relative phase standard deviation. Reporting research in an effective manner facilitates the dissemination of knowledge and subsequent advancement of science.