Pros and cons of group averaging in studies of stochastic rodent locomotion

We analyze spontaneous walks of rodent pups in a novel environment. Individual speed distributions follow a two-component Rayleigh distribution, reflecting bimodal locomotion (progressions and lingering), while the ensemble- averaged distribution simplifies to a single exponential decay with a slower high-speed decay rate. This discrepancy arises from population heterogeneity in characteristic speeds. We propose a stochastic differential equation (SDE) for the averaged rodent, which in some cases can be derived by averaging the individual SDE. The SDE's reveal a striking mechanistic difference in the nature of individual and averaged motion: individual rodents exhibit viscous friction (deceleration proportional to speed), whereas the ensemble behaves as if subject to dry friction (constant deceleration). We propose an algorithm of creating a "reference" rodent, which retains the shape of individual distributions but utilizes averaged scale parameters. It can be a useful personalized comparison tool to extract the information on the locomotor modes, e.g. in biomedical contexts.