Causal Inference Toolkit

Production-ready Python library & Streamlit web dashboard wrapping DoWhy/EconML with sensitivity analysis, quasi-experiments, A/B testing, and executive HTML report generation.

PythonDoWhyEconMLStreamlitPytestPandasScikit-Learn

View Source Live Demo

The Problem

Observational studies and A/B tests often lack rigorous sensitivity analysis, unified estimation workflows, and executive-ready reporting, leading to unmitigated confounding bias and slow decision cycles.

The Solution

Built an end-to-end Python library and Streamlit dashboard combining SOTA causal estimators, quasi-experimental methods, sensitivity bounds, and one-click HTML executive report exports.

Architecture

Modular Python architecture with core causal abstractions, wrapper layer over DoWhy/EconML, dedicated analysis engines (Sensitivity, DiD, SCM, A/B testing, Uplift), Streamlit UI app, Typer CLI, and Jinja2-based executive HTML report generator.

Key Features

Unified DoWhy & EconML estimation pipeline (IPW, Matching, Doubly Robust, Double ML, Causal Forest)

Quasi-experiments engine: Synthetic Control Method (SCM) & Difference-in-Differences (DiD) event studies

Multi-layer sensitivity analysis (Rosenbaum bounds, Cinelli-Hazlett, E-values, TIPS curves)

Interactive Streamlit dashboard & automated standalone HTML executive report generator

Deep Dive

Causal Inference Toolkit is an enterprise-grade Python package and interactive Streamlit web dashboard that unifies observational causal inference, quasi-experiments (Synthetic Control, Difference-in-Differences), sensitivity analysis (Cinelli-Hazlett, E-values, Rosenbaum bounds), A/B testing (Frequentist/Bayesian/Sequential), and uplift metalearners (T/S/X/R/DR-learners). Features automatic executive HTML report generation, CLI tools, and visualization pipelines for causal graphs and covariate balance.

Future Roadmap

1Add neural causal estimators (TARNet, Dragonnet)

2Implement continuous treatment effect estimation

3Expand automated DAG discovery algorithms

Deep Dive