pgmpy/pgmpy

Python Toolkit for Causal and Probabilistic Reasoning

What it solves

pgmpy is a library for causal and probabilistic reasoning using graphical models. It allows users to build, learn, and perform inference on models that represent complex dependencies between variables, enabling the discovery of causal relationships and the calculation of probabilities based on observed evidence.

How it works

The library implements various data structures for graphical models, including Bayesian Networks, Dynamic Bayesian Networks, and Structural Equation Models. It provides a modular, composable API for tasks such as:

  • Causal Discovery: Learning the model structure from data or expert knowledge.
  • Parameter Learning: Estimating conditional probability distributions from observed data.
  • Probabilistic Inference: Computing posterior distributions based on evidence.
  • Causal Inference: Computing interventional distributions.
  • Simulations: Generating synthetic data under specific interventions or evidence.

Who it’s for

It is designed for researchers and developers who need to perform probabilistic graphical modeling, causal discovery, and causal inference in Python, with compatibility for scikit-learn pipelines.

Highlights

  • Diverse Model Support: Supports DAGs, PDAGs, MAGs, PAGs, and Linear Gaussian Bayesian Networks.
  • Causal Tooling: Integrated tools for causal discovery, identification, and validation.
  • Causal and Probabilistic Inference: Ability to compute both posterior and interventional distributions.
  • Scikit-learn Compatible: Modular API that can be integrated into standard ML pipelines.
  • Flexible Data Handling: Supports discrete, linear Gaussian, and mixture data with arbitrary relationships via Functional Bayesian Networks.

Related

  • Project
    SheffieldML/GPy

    A Gaussian processes framework in Python for implementing and performing regression and other GP-based modeling tasks.

  • Project
    cdt15/lingam

    A Python library for causal discovery that estimates linear non-Gaussian acyclic models to determine the causal relationships between variables.

  • Project
    py-why/dowhy

    DoWhy is a Python library for end‑to‑end causal inference. It lets you define a causal graph, automatically identifies the causal effect, estimates it with a variety of statistical methods, and runs refutation tests to check robustness. The library also supports graphical causal models for root‑cause analysis, counterfactuals, and interventional sampling. Install via `pip install dowhy` and follow the four‑step workflow (model → identify → estimate → refute).

  • Project
    erdogant/bnlearn

    bnlearn is a Python library for learning, parameterising, and querying Bayesian networks. It supports structure discovery (various scores and search methods), CPT estimation, causal inference (do‑calculus), synthetic data generation, and visualisation, all wrapped in a simple API.

  • Project
    pymc-labs/CausalPy

    CausalPy is a Python library for Bayesian‑first causal inference in quasi‑experimental settings. It implements methods such as difference‑in‑differences, synthetic control, regression discontinuity, and instrumental variables, offering posterior‑based effect summaries, credible intervals, ROPE analysis, and publication‑quality plots. The package builds on PyMC 6, works with Python 3.12+, and provides both Bayesian and OLS options for estimation.