The Effects of Examination Methods on Describing the Scale-Free Nature of Human Locomotor Activity

Although human dynamics exhibit scale-independent properties, different results have emerged from the statistical analyses of various actigraphic or mobility datasets regarding the characterization of their heavy-tailed distributions. By applying approaches from significant studies of this nature of human locomotor activity on our multi-day long acceleration dataset, we examine how the non-standardized activity determination methods and the separation of the resulting activity data into active and passive periods affect the conclusions. We also show that the statistical or model fitting approaches commonly used to study these periods not only influence the results but are also highly sensitive to the actigraphic methodology employed. In contrast, the differently derived activity data — and even the underlying acceleration signals — measured on the wrist, follow universal 1/f noise-based spectral characteristics that have also been previously observed in the context of human spatial motion, thus offering the possibility of a more robust analysis of scale-independence.