Do contacts over distance follow a power-law distribution? Estimation of the social contact distance kernel

Abstract

Do contacts over distance show a multimodal form, with a peak of contacts close to home and a second peak further away from home, or is a powerlaw form su fficient? By using data from our social contact study, we were able to test this hypothesis. We exploited various distributions for the contacts at a certain distance, e.g. Poisson, Negative Binomial, . . . , and incorporated random eff ects to account for the clustering of contacts within participants. Various forms of the underlying distribution were tested by integrating their information into the observed categories. The preliminary results support a Weibull form for the distribution of contacts over distance, however subtle di fferences are present when diff erentiating by the participant's age and by week or weekend days.