liveBook · Manning

Minimal d-separating set

As explained in Causal AI

A minimal d-separating set is a d-separating set whose blocking property is not represented by a smaller relevant d-separating set.

A minimal d-separating set blocks every d-connecting path between two nodes while avoiding unnecessary members: removing any relevant member would fail to preserve the separation. Larger valid sets may contain a minimal set, but algorithms should focus on the minimal alternatives. section 4.2.1

Find candidate sets by checking paths and the effect of colliders:

  • Enumerate the d-connecting paths between the target nodes. section 4.2.2
  • Block ordinary path nodes, remembering that conditioning on a collider can open a path that was previously d-separated. section 4.2.1
  • Retain only sets for which no relevant member can be removed without reopening a path. section 4.2.1

A collider can require an extra blocker

For U and V, M blocks U ← M → V and the other listed d-connecting paths. But conditioning on M opens U ← I → M ← J → V, so either I or J must also be included. The minimal sets are {I, M} and {J, M}. section 4.2.2

Figure 4.15 Blocking with M will block the path U ← M → V but would d-connect the path U ← I → M ← J → V because M is a collider between I and J. So we need to additionally block on either I or J to d-separate U ← I → M ← J → V.
figure

Inspect the path diagram showing why blocking at collider M requires an additional blocker at I or J. section 4.2.2

How Causal AI covers it

How deep it goes
Works through path enumeration, collider effects, subset comparison, and concrete separating sets.
Examples
Uses paths through U, V, I, M, and J to derive the minimal sets {I, M} and {J, M}.
What it assumes
Builds from individual-node path intuition before generalizing d-separation to sets of nodes.
What you can do after
Apply path checks to identify nonredundant conditioning sets and detect paths opened by collider conditioning.
How it connects
Connects d-separation with d-connection, colliders, conditioning sets, node sets, and do-calculus.

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Causal AI. The section links point to the parts of the book it draws on.