Network Analysis Explained: How to Make, Read and Report It
Network analysis describes a system as nodes (units) connected by edges (relationships) and then quantifies its structure: who is most connected, how dense the network is, and how information or influence could travel. This post focuses on the metrics. For how to draw and style a network, see the separate network plot guide. The existing StatFAS Mantel network post concerns the integrated Mantel tool and is a different topic.
Core concepts
For an undirected, unweighted network with n nodes and m edges: degree ki is the number of edges at node i (Σki = 2m); density = 2m / [n(n − 1)]; the shortest path between two nodes is the fewest edges linking them; closeness of node i = (n − 1) / Σ d(i, j); betweenness (Freeman, 1977) counts how often a node lies on shortest paths between other pairs; clustering measures how often a node's neighbours are themselves connected (Watts & Strogatz, 1998).
Original example (illustrative)
Five researchers A–E with collaboration edges A–B, A–C, B–C, C–D and D–E (n = 5, m = 5).
| Node | Degree | Closeness | Betweenness | Local clustering |
|---|---|---|---|---|
| A | 2 | .57 | 0 | 1.00 |
| B | 2 | .57 | 0 | 1.00 |
| C | 3 | .80 | 4 | .33 |
| D | 2 | .67 | 3 | 0 |
| E | 1 | .44 | 0 | 0 |
Degrees sum to 10 = 2m. Density = 10/20 = .50. The sum of all pairwise shortest-path lengths is 17 over 10 pairs, so the average path length is 1.7 and the diameter is 3 (A–E and B–E). C's betweenness of 4 comes from the pairs A–D, A–E, B–D and B–E, all of which must pass through C. D's value of 3 comes from A–E, B–E and C–E. Global transitivity = 3 × 1 triangle / 6 connected triples = .50. The network is held together by C and D: removing C disconnects A and B from D and E.
This is a toy network. With real data, metrics depend on how edges were defined (a correlation threshold, a survey tie, a co-authorship), so describe that choice explicitly.
How to make it in StatFAS
Open the Network Analysis Calculator. It sits alongside the PCA calculator in StatFAS's tool list.
Reporting
The collaboration network comprised 5 nodes and 5 edges (density = .50, average path length = 1.70, diameter = 3). Node C had the highest degree (3) and betweenness (4), indicating a bridging role.
Common mistakes and alternatives
Comparing metrics across networks of different size without normalisation, over-interpreting centrality in tiny or sampled networks, ignoring edge weights and direction, building networks from arbitrary correlation cut-offs, and treating centrality as importance regardless of context. Alternatives: community detection, regression-based network models and ERGMs (Newman, 2018).
Conclusion
Network metrics summarise structure, but only a clear definition of nodes and edges makes them scientifically meaningful.
1.
statfas-network-analysis-interface.png, alt: "StatFAS Network Analysis Calculator interface"2.
statfas-network-analysis-example-data.png, alt: "Example data entered in the StatFAS Network Analysis Calculator"3.
statfas-network-analysis-settings.png, alt: "Network Analysis Calculator settings in StatFAS"4.
statfas-network-analysis-result.png, alt: "Network Analysis Calculator result generated in StatFAS"References
- Freeman, L. C. (1977). A set of measures of centrality based on betweenness. Sociometry, 40(1), 35–41.
- Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393, 440–442.
- Newman, M. E. J. (2018). Networks (2nd ed.). Oxford University Press.