PCAlgorithm<T>

Models & Types · AiDotNet.CausalDiscovery.ConstraintBased

PC Algorithm — constraint-based causal discovery using conditional independence tests.

Inheritance: CausalDiscoveryBase<T> → ConstraintBasedBase<T> → PCAlgorithm<T>

For Beginners

PC figures out which variables cause which by testing: “Are X and Y still related after we account for other variables?” If X and Y become unrelated (conditionally independent) given some set of variables, the edge between them is removed. Then remaining edges are oriented to form a DAG.

How It Works

The PC (Peter-Clark) algorithm learns a causal graph by:

  • Starting with a complete undirected graph
  • Removing edges between conditionally independent variable pairs
  • Orienting edges using v-structures and orientation rules

Reference: Spirtes, Glymour, and Scheines (2000), “Causation, Prediction, and Search”.

Constructors

Constructor Summary
PCAlgorithm(CausalDiscoveryOptions options) Initializes the PC algorithm with optional configuration.

Properties

Property Summary
bool SupportsLatentConfounders { get; }
bool SupportsNonlinear { get; }
string Name { get; }