Network of DSAMs that are positively correlated.

<p>(a) Complex network representation of the positive connections among crime and property DSAMs. Each node is a crime or property type and the connection between two nodes occurs whenever there is a statistically significant correlation between their DSAMs (based on bootstrapping the Pearson correlation and 99% confidence). Each connection is weighted by the Pearson correlation coefficient and the thickness of the edges are proportional to the connection weight. Node sizes are proportional to their degrees and the color code also refers to node degree. A modular structure composed of two modules (one with all property metrics and a second with all crime metrics) is identified by maximizing the network modularity (yielding <i>M</i> = 0.47 for the original network and 〈<i>M</i>_rand〉 = 0.12 ± 0.01 for a set of randomizations of the original network). Edges highlighted in blue are ones connecting the two modules. (b) Characterization of nodes based on the within-module connectivity (<i>W</i>) and participation coefficient (<i>P</i>). Each dot in the <i>W</i>-<i>P</i> plane corresponds to a crime or property type. All nodes are classified as ultraperipheral (<i>R</i>1) or peripheral (<i>R</i>2); in particular, the majority of nodes has zero participation coefficient (that is, has only within-module links) and only the six nodes in the <i>R</i>2 region have between modules connections. (c) Modular structure of the sub-graph related to the crime metrics. For this case, two modules (colored in purple and green) are found by maximizing the network modularity (<i>M</i> = 0.14 and 〈<i>M</i>_rand〉 = 0.06 ± 0.01). (d) Role discrimination of crime nodes by the <i>W</i>-<i>P</i> plane. We note that all nodes are in the peripheral region (<i>R</i>2). Drugs, order, and anti-social behavior (ASB) crime types are the most peripheral; robbery and burglary have the largest <i>P</i>, and criminal damage and arson (CD and A) has the largest <i>W</i>.