Network robustness

This folder contains data and R-codes to assess social network robustness. We used two approaches to do this: we first assessed, for each data collection method and for each social behaviour, the variation in mean value of all three individual-level network measures as well as the variation of all three network-level measures over time with monthly increases of data collected. We expected the curves to become progressively “flatter” because, if the networks were becoming more and more stable over time, monthly variation in network measures would become smaller and smaller as observers recorded fewer and fewer new edges between nodes. Second, we used bootstrapping to estimate network uncertainty, which reflects the (un)certainty with which network metrics were estimated. For each monthly data and for each social behaviour examined, the identity of the recipient was randomly reshuffled and social network metrics were re-calculated. This procedure was repeated 1000 times, eventually generating a distribution of possible values. From this distribution, we extracted the 95% confidence interval and subtracted the maximum and minimum value of this range in order to calculate the uncertainty index. We then assessed, for both sampling method, the monthly variation of this uncertainty index, expecting this value to decline as more and more observations were recorded and networks would become more certain.