Methodological issues associated with the mean value of repeated laboratory running measurements

Abstract

Many studies examining the biomechanics of running are conducted in laboratories, where spatial and instrumental limitations often result in protocols containing repeated running trials. Data from these repeated trials are then usually pooled into mean values to represent the roll-over characteristic for each subject. The purpose of the present study was to analyse the minimum number of trials necessary to achieve adequate validity for these protocols. We used an empirical approach, investigating 14 subjects who performed 100 running trials in a laboratory on two test days. Ground reaction forces were recorded and standard loading variables were calculated from the time series data. The convergence of cumulated mean values was analysed to verify basic assumptions about the random characteristic of repeated measurements. Analyses of differences within sessions and the root mean square error (RMSE) for repeated measurements were used to assess the absolute reliability of the repeated trials. Depending on the variable analysed, 29–57% of the subjects’ cumulated mean values demonstrated convergence. The minimum number of trials to achieve convergence was found to be between 58 and 68. The maximum differences were found to be 33% and 38% within and between days, respectively. In conclusion, the random characteristic of repeated measurements could not be confirmed, and the mean did not serve as a valid estimator for the true value of a subject. As a consequence, we suggest exploring alternative possibilities to overcome basic methodological issues of laboratory protocols in running and to reconsider the results of many previously published studies on the biomechanics of running.

Keywords:

• Variability
• running
• laboratory conditions
• kinetics, mean value

Notes

¹ We do not focus on systematic offset in this paper because it does not produce additional variability around the mean of the repeated measurements. Rather, it reflects a systematic offset between the true quantity and the measurement result, and does not affect the minimum number of trials necessary.