Estimation of Global Network Statistics from Incomplete Data [arxiv]

Catherine A. Bliss, Christopher M. Danforth, Peter Sheridan Dodds

Figure 1:

Figure 1

Node induced subnetwork on randomly sampled nodes.(a.) The true network is sampled by randomly selecting nodes (red). (b.) The node induced subnetwork consists of sampled nodes and edges whose endpoints both lie in the collection of sampled nodes.

Figure 2:

Figure 2

Failed link subnetwork. Hidden or missing links are depicted in grey. All nodes remain in the subnetwork and only visible or sampled links remain.

Figure 3:

Figure 3

Subnetwork generated from sampled links. (a) A network is sampled by randomly selecting links shown in red. (b) The subnetwork consists of all sampled links and only nodes which are incident with the sampled links. In this type of sam- pling, no nodes of degree zero are included in the network. Large degree nodes are more likely to be included in the subnetwork.

Figure 4:

Figure 4

Weighted subnetwork generated from sampled interactions. (a.) An unsampled weighted network consists of nodes, links and weights representing the number of interactions represented by the link. (b.) Sampling by interacting produces a subsample whereby links are included in the subsample only if at least one interaction has been sampled. The subnetwork is the induced subgraph on these links with wi≥1.