About
Tetrad is an open-source platform for causal discovery, simulation, and causal model analysis developed at Carnegie Mellon University and the University of Pittsburgh. Originating in the philosophy of science and machine learning communities, the project has grown over several decades into a comprehensive environment for learning, evaluating, and reasoning with graphical causal models.
This manual provides consolidated, up-to-date documentation for Tetrad’s algorithms, graphical model representations, data formats, effect-estimation tools, and user interfaces. It replaces and expands the older “classic” HTML manual, incorporating modern algorithms, updated descriptions, and extensive cross-referencing.
📚 Project Background
Tetrad supports a full ecosystem of causal modeling tools, including:
DAG, CPDAG, MAG, and PAG causal graphs
constraint-based, score-based, and hybrid search algorithms
methods for learning with latent variables and selection bias
effect-size estimation, adjustment sets, and path blocking
nonlinear and non-Gaussian independence tests and scores
simulation frameworks for linear, nonlinear, and neural-network–based SEMs
extensive GUI tools and programmatic interfaces
In addition to the Java/GUI version, Tetrad includes modern APIs:
Py-Tetrad — Python interface (with its own 2023 software paper)
RPy-Tetrad — R interface via rpy2
The core Java library, used both in the GUI and headless workflows.
Development is ongoing, with new algorithms and methods added as time and resources allow. Community suggestions and contributions are welcome.
For a broader historical overview of the project, see:
👥 Contributors
Tetrad has benefited from the work of many collaborators, including developers, research scientists, graduate students, and external contributors.
A maintained list appears here:
📄 Papers and Books
A curated bibliography of foundational publications, software papers, theoretical works, and algorithm references relevant to Tetrad and its tools:
This list includes classic references alongside modern work on hybrid, nonlinear, and high-dimensional causal discovery.
📬 Questions or Suggestions?
Suggestions for improving the manual are welcome.
Please open an issue or discussion on our GitHub repository:
👉 https://github.com/cmu-phil/tetrad-manual-rtd