Figure 8

Assortativity coefficient. The plot shows the influence of the individual regions on the overall assortativity. For the network of families (circles) calculation is done using the region of origin of the families, and for the network of kins (squares), using the birth region of individuals. Here we estimate the influence of each region by using jackknife resampling [33]. For a given network, first the overall assortativity coefficient (a) is calculated using information on all the nodes (see Appendices). Then for a given region, all the nodes belonging to the region are removed, and the assortativity coefficient is recalculated (\(a^{*}\)). This calculation is repeated for all the regions, and the mean \(\langle a^{*}\rangle\) is found. The difference \(a^{*}-\langle a^{*}\rangle\) for a particular region (denoted on the horizontal axis) indicates the effect of removal of the nodes from the network. For example, a value of \(a^{*}-\langle a^{*}\rangle\) being significantly greater than zero, indicates that the assortativity increases when the nodes are removed implying that the nodes themselves participate in disassortative mixing. For the kin network, the error bars in the quantity \(a^{*}\) are shown. For the network of families, we observe a clear correspondence of the values of \(a^{*}-\langle a^{*}\rangle\) with the values in case of the kin network. However, the errors are much larger in magnitude (due to a lower link density) and therefore, are not shown in the figure