Analyses

Network centrality

Summarise a graph of nodes and the links between them.

Edit on GitHub

When to use it

Use this when the rows are relationships, not people-by-variable scores. An edge list has a source column and a target column. An adjacency block is a square set of columns, one per node. The result is who is central, how dense the graph is, and how it splits into communities.

Assumptions

Give either an edge list or adjacency columns, not a vague mix. Node names are read as text. An optional weight column is numeric. A graph with no nodes is an error. Centrality uses an undirected reading of the links.

Running it in Tensr

Analyze → Network → Centrality. In chat: “Centrality for this edge list.”
For an edge list, set source and target. Weight is optional.
Or set adjacency columns for a square matrix and leave source and target empty.

Options

Prop

Type

Reading the output

A 12-person class. An edge is a study link, kept when a seeded draw fell under 0.28, with a weight from 1 to 4. Centrality is descriptive.

NodeDegreeBetweennessClosenessEigenvector
s10.2730.0910.5240.168
s20.4550.1270.6110.434
s40.2730.0360.440.127
s100.4550.2180.6110.297
s30.3640.0910.5790.338
s70.2730.1360.550.236
s110.18200.4580.164
s120.3640.0550.5790.437

Network with 12 nodes and 22 edges. Network with 12 nodes and 22 edges. Metrics: Nodes = 12; Edges = 22; Density = 0.333; Components = 1; Modularity = 0.251.

Reporting (APA 7)

Network with 12 nodes and 22 edges. Report the estimate in the table. This procedure is not summarised by one p-value.

Coming from SPSS

The product menu is Analyze → Network → Centrality. SPSS does not have this dialog. The spring layout is a scatter of nodes, with a fixed seed so the picture does not jump between runs.