Statistical properties of human activity and criticality in active behavior

Abstract

We study the statistical properties of human activities by analyzing actigraphy data that monitor wrist movements of normal subjects and patients with bipolar II disorder. Probing the distribution of sizes, lifetimes, and quiet times for active bouts, we identify power-law behavior and detect a significant difference between the two groups in the scaling exponent of the quiet-time distribution. Scale invariance in active bouts is further examined through the dynamical exponent and scale transformation while temporal correlations of the activities are explored by means of the detrended fluctuation analysis. In both normal subjects and patients, long-time correlations are apparent in wake periods which bear most active bouts. These correlations are, albeit relatively weak, related to the amount of active bouts during sleep. Our findings suggest that human movements in daily life are regulated to (near-)criticality.