Figure 2

Classifying communities as non-segregated and highly segregated. (A) and (B) are examples of ego-networks of co-authorships in 2010 of non-segregated and highly segregated communities, in light blue and red, respectively. Ego-networks are sub-graphs induced by the connections between central nodes, i.e., ego (colored nodes belonging to the selected community) and their one-step neighbours, i.e., alters (dark grey nodes belonging to other communities connected to the colored community). Edges inside the communities have the color of the nodes, while links across communities are in light grey. (C) shows the probability density function (PDF) of the spectral segregation index (SSI) for 2010. The plot is divided into three categories that denote non-segregated (light blue), moderately (grey), and highly segregated (light red) communities. The complete procedure is in Section S4. For the distribution, we use a Gaussian kernel density estimation with the “rule of thumb” for the bandwidth selection [36]