. 2015 Sep 11;10(9):e0137248. doi: 10.1371/journal.pone.0137248

Table 2. Network modularity.

	Communities	Modularity	Towns in modules			Cramer’s V
			Min	Max	Mean	w _ij ⁽²⁾–1	w _ij ⁽²⁾–2	w _ij ⁽²⁾–3	w _ij ⁽²⁾–4
w _ij ⁽¹⁾	5	0.209	254	720	511.6
w _ij ⁽²⁾–1	14	0.373	72	285	182.7	1
w _ij ⁽²⁾–2	15	0.372	72	285	170.5	0.983	1
w _ij ⁽²⁾–3	13	0.372	106	285	196.7	0.968	0.982	1
w _ij ⁽²⁾–4	14	0.371	72	285	182.7	0.954	0.965	0.953	1
w _ij ⁽²⁾–5	15	0.372	72	270	170.5	0.984	0.960	0.981	0.963

Few large communities are identified in the $w_{i j}^{(1)}$ network compared to the $w_{i j}^{(2)}$ network. The five runs of the Louvain method finds exactly the same community structure in the $w_{i j}^{(1)}$ network, therefore we report the result of only one run. The community finding algorithm produced distinct community structures in the $w_{i j}^{(2)}$ network; therefore we report all five runs (the number after the hyphen denotes the sequence of the run). The pair-wise Cramer’s V index in $w_{i j}^{(2)}$ network is always above 0.95.