Constraint-Based Causal Discovery Suite

A four-package Python suite for constraint-based causal discovery — simulation, conditional independence testing, structure learning, and metric / visualisation. Each package is independent and stands on its own; cross-package interoperability is via structural Protocols, not imports.

Package documentation

dagsampler

Configurable DAG / SCM simulator producing synthetic mixed-type data and an optional CI oracle.

cbcd

Constraint-based causal discovery algorithms: PC, FCI, RFCI, anytime-FCI, PCMCI.

citk

Conditional independence test toolkit: FisherZ and Spearman native; KCI / CMIknn / RegressionCI / GCM and others via optional extras.

bnm

DAG / CPDAG / PAG comparison metrics and visualisation: SHD, HD, F1, SID, per-Markov-blanket comparisons.

Source: github.com/averinpa/constraint-based-causal-discovery-suite · MIT licensed