Documentation
Complete guide to using CausalLab for AI-powered causal inference research
What is CausalLab?
CausalLab is an AI-powered research platform designed for social scientists, economists, and management researchers. It automates complex causal inference workflows, from discovering causal relationships in data to simulating policy interventions.
- Discover hidden causal relationships from complex data automatically
- Get AI-recommended identification strategies and control variables
- Automatically decompose effects into direct and indirect pathways
- Simulate counterfactual policies before implementation
Projects vs Templates
Understanding the distinction between Projects and Templates is fundamental to using CausalLab effectively.
Projects
Active WorkWhat: Live research instances with real data
Contains: Your dataset, analysis results, saved progress
Use for: Conducting actual research studies
Templates
Reusable BlueprintWhat: Pre-configured research designs
Contains: Methodology, variable definitions, no data
Use for: Starting similar studies quickly
Quick Start Guide
Get started with CausalLab in three simple steps:
Upload Your Data
Upload CSV, Excel, or STATA files. Connect to databases like Supabase. Define your treatment, outcome, and control variables.
Run Analysis
Progress through the four modules: Causal Discovery → Identification → Mechanism Analysis → Policy Simulation. Export results at any stage.
Causal Discovery Lab
The Causal Discovery Lab uses AI algorithms to automatically identify potential causal relationships in your data, generating visual causal graphs.
Key Features
- Algorithm Selection: Choose from PC, GES, FCI, or hybrid methods based on your data structure
- Interactive Graph: Visualize causal relationships with directed edges showing effect strength
- AI Insights: Get literature comparisons and novel relationship suggestions
Identification Strategy Optimizer
Automatically recommends econometric identification strategies based on your research design and data structure.
Supported Methods
Mechanism Path Analyzer
Decompose total effects into direct and indirect pathways, identifying the mechanisms through which treatment affects outcomes.
Example: Carbon Trading Effect Decomposition
Policy Simulation Sandbox
Simulate counterfactual scenarios to predict policy impacts before implementation, with uncertainty quantification.
Simulation Methods
- Neural ODE for continuous dynamics
- PySINDy for discovering governing equations
- Agent-Based Models for heterogeneous agents
- Monte Carlo for uncertainty quantification
Template-Based Research Workflow
Leverage pre-configured templates to accelerate your research process:
Browse Template Library
Explore pre-built templates in /examples or your saved templates in /templates
Load Template
Click "Load Template" to create a new project with pre-filled research design
Customize & Upload Data
Adjust variables if needed and upload your actual dataset
Run Analysis Pipeline
Progress through all four modules with methodology already configured
Exporting Results
Export your analysis results in multiple formats for publication and presentation:
PDF Report
Publication-ready comprehensive report with all tables, figures, and methodology
Excel Data
Raw regression results, effect decomposition, and simulation outputs
STATA/R Code
Reproducible analysis scripts for external validation
Causal Graph
High-resolution causal graph visualization in PNG/SVG format