A Method to Stop Analyzing Random Error and Start Analyzing Differential Responders to Exercise

Abstract

It is commonly stated that individuals respond differently to exercise even when the same exercise intervention is performed. This has led many researchers to conduct exercise interventions and subsequently categorize individuals into different responder categories to determine what causes individuals to respond differently. Some methods by which differential responders are categorized include percentile ranks, standard deviations from the mean, and cluster analyses. Notably, each of these methods will result in the presence of differential responders even in the absence of an exercise intervention, indicating that individuals may be categorized based on the presence of random error as opposed to true differences in the exercise response. Here we propose a method by which differential responders can be classified after accounting for the presence of random error that is quantified from a time-matched control group. Individuals who exceed random error from the mean response of the intervention group can be confidently labelled as high and low responders. Importantly, the number of differential responders will be proportional to the ratio of variance in the exercise and control groups. We provide easy-to-follow steps and examples to demonstrate how this technique can identify differential responders to exercise. We also detail the flaws in other classification methods by demonstrating the number of differential responders who would have been classified using the same data set. Our hope is that this method will help to avoid misclassifying individuals based on random error and, in turn, increase the replicability of differential responder studies.