Fig. 7

Results regarding users whose network centralities or physical activities change with time. a The number of significantly-changing users for each of the 10 centralities individually (ECC to SMS_F) or by at least one of them (NET_T), for physical activity (PA_T), and for the union of NET_T and PA_T (the last bar). Red and blue portions of the given bar indicate users that are positively and negatively correlated with time, respectively. b The number of users found by exactly k centralities (k = 1, 2,..., 10). c The number of users that are unique to the given centrality, i.e., for a given user, only for this centrality, the user’s centrality values significantly change with time. d Pairwise overlaps between significantly-changing users predicted by the different centralities or physical activity. An overlap is measured as Jaccard index, and the resulting overlap size is indicated by the given color. The significance of the overlaps is computed by using the hypergeometric test. *, **, and *** indicate statistical significance at p-value thresholds of 0.05, 0.01, and 0.001, respectively