Figure 3
The closeness centrality of MLVA profiles is compared to the prevalence of their clusters. The clusters are obtained by either the overlapping (top row) or the partitioning (bottom row) algorithms. For overlapping clusters (threshold distance five), the individual centrality of each node is plotted against the logarithm of average prevalence of their cluster. For partitioning clustering (with 21 mutually exclusive clusters), the average centrality of each cluster is plotted against the logarithm of average prevalence of that cluster. The size of each circle (left subfigures) is set in proportion to the size of the corresponding cluster. Solid lines trace binned averages, using: 50 equal size bins in the range between 0.0233 and 0.0831 for the top sub-figure (i.e., individual node centrality); and 10 equal size bins in the range between 0.0233 and 0.0710 for the bottom sub-figure (i.e., average cluster centrality). Dashed lines trace the corresponding standard deviations above the means. The scattergrams are coloured by the pairwise L1-norm between nodes (top left) or clusters (bottom left). That is, for the overlapping approach (top left), each node is coloured according to distance to other focal MLVAs, whereas in the partitioning approach (bottom left), the average pairwise distance between all nodes in two clusters is used. The left subfigures show how far these clusters are, in terms of these distances, from the most prevalent cluster, with a clear colour gradient. The right subfigures visualise the nodes within the most prevalent cluster as opaque in each network, with all other nodes semi-transparent. The size of each node (right subfigures) is set in proportion to the prevalence of the corresponding MLVA profile.
