Statistical Methods
Find Disconnected Groups in an Undirected Network
Define the task
Connected components identify groups linked by at least one path. They are useful when checking whether a comparison network or processing dependency map splits into disconnected parts. First define what an edge means and whether direction matters. This tool accepts a symmetric binary adjacency matrix and treats the graph as undirected. It does not estimate treatment effects, assess evidence quality or infer that two connected nodes are directly comparable without additional assumptions.
Carry out the check
Check that rows and columns use the same node order. Every off-diagonal one must have a matching one across the diagonal; diagonal entries must be zero. Starting from an unvisited node, follow edges until no reachable unvisited nodes remain, then begin another component. Isolated nodes each form their own component. Component labels are bookkeeping, not ranks. Directed networks need an explicit weak-versus-strong connectivity definition and are rejected by this interface rather than silently symmetrized.
Worked hypothetical example
For a hypothetical four-node graph with edges 0–1 and 1–2, nodes 0,1,2 share one component while node 3 is isolated. There are two components with sizes three and one. Removing edge 1–2 increases the count to three. Preserve the adjacency rule and node labels with this result. A connected graph can still contain fragile single-edge bridges, and connectivity alone does not validate transitivity, consistency or the scientific meaning of indirect comparisons.
Method checked 21 November 2026 against Primary method documentation: connected_components.