Accounting for individual responses is an important issue in studies of the effects of physical training and other interventions on performance and health. Individual responses can be understood as variation in change scores additional to the variation arising from error of measurement in both parallel -group and crossover randomized controlled trials. Estimation of this additional variation with adequate precision often requires impractically large sample sizes, and there are computational challenges with quantifying the extent to which each individual's response can be attributed to the intervention. However, subject characteristics that might explain individual responses should be measured and included in the usual analyses as modifying covariates, because their effects can be estimated often with sufficient precision to allow identification of the kind of individuals who are likely to be positive, trivial, or negative responders. A sample size four times larger than that required to estimate the mean effect of the intervention is a potential limitation for assessing such effects, so researchers should reduce the requisite sample size by using Bayesian inference with a dispersed uniform prior (magnitude-based inference), by studying only one population subgroup (e.g., females), and by opting for various strategies to reduce error of measurement, including choice of the most reliable measure for the dependent variable, inclusion of habituation trials where appropriate, and averaging of repeated trials before and/or after the control and experimental treatments. The analyses should include provision for the extra variation arising from individual responses to ensure the effects of the intervention and modifiers are estimated correctly.
© Copyright 2018 Sportscience. AUT University. Alle Rechte vorbehalten.