High-performance implementation of Shpitser's ID algorithm for causal identification, written in Rust. While working on this project, I found a bug in Microsoft Research's DoWhy library that was invalidating ~50% of significance tests. My implementation is more general and 60% faster than Microsoft's version.

Overview

The core is Shpitser's ID algorithm, a fundamental algorithm in causal inference for determining whether a causal effect can be identified from observational data given a causal graph. There were a couple of open source implementations available when I started the project, but they suffered from a mixture of performance and maintenance issues. So, I implemented the core algorithm in Rust and then put it in a Python wrapper.

Key Achievements

• Bug Discovery: Found a critical bug in Microsoft Research's DoWhy library that was invalidating approximately 50% of significance tests
• Performance: Achieved 60% faster execution compared to Microsoft's implementation
• Generality: Built a more general implementation that handles a wider range of causal structures
• Language: Implemented in Rust for performance, with Python bindings for ease of use

Technical Details

Causal graphical models are a mathematical framework for reasoning about cause and effect relationships in data. The ID algorithm determines whether it's possible to compute causal effects from observational data, given assumptions encoded in a directed acyclic graph (DAG).

This project combines the performance benefits of Rust with the accessibility of Python, making it suitable for both research and production use cases in causal inference.