Tutorials

Welcome to the CausalFM Toolkit tutorials! These step-by-step guides will help you get started with causal inference using foundation models.

Tutorial Series

• Tutorial 1: Basics
  • Learning Objectives
  • What are Foundation Models for Causal Inference?
  • Key Concepts
  • Causal Settings in CausalFM
  • Basic Workflow
  • Your First CausalFM Script
  • Key Takeaways
  • Next Steps
  • Questions?
• Tutorial 2: Data Generation
  • Coming Soon
  • Quick Reference
  • Next Tutorial
• Tutorial 3: Training Models
  • Coming Soon
  • Quick Reference
  • Next Tutorial
• Tutorial 4: Model Evaluation
  • Coming Soon
  • Quick Reference
  • Congratulations!
  • Next Steps

Tutorial Overview

Tutorial 1: Basics

Learn the fundamentals of CausalFM:

• Understanding the foundation model approach

• Key concepts: PFNs, in-context learning, CATE estimation

• Basic workflow from data to predictions

Tutorial 1: Basics

Tutorial 2: Data Generation

Master data generation for different settings:

• Standard CATE data generation

• Instrumental variables data

• Front-door adjustment data

• Understanding DAG-structured SCMs

Tutorial 2: Data Generation

Tutorial 3: Training Models

Learn how to train your own models:

• Configuring training runs

• Monitoring training progress

• Saving and loading checkpoints

• Training for different causal settings

Tutorial 3: Training Models

Tutorial 4: Model Evaluation

Evaluate your models effectively:

• Computing causal inference metrics

• Uncertainty quantification

• Visualization techniques

• Comparing multiple models

Tutorial 4: Model Evaluation

Prerequisites

Before starting these tutorials, make sure you have:

• ✅ Installed CausalFM Toolkit (see Installation)

• ✅ Basic Python knowledge

• ✅ Familiarity with PyTorch (helpful but not required)

• ✅ Understanding of causal inference concepts (helpful but not required)

What You’ll Learn

By completing these tutorials, you will be able to:

1. Generate synthetic causal datasets

2. Train foundation models for causal inference

3. Make predictions with pretrained models

4. Evaluate model performance

5. Apply CausalFM to your own causal inference problems

Example Notebooks

For hands-on examples, check out the Jupyter notebooks in the evaluation/notebook/ directory:

• test_standard_cate.ipynb - Standard CATE estimation

• test_iv_binary.ipynb - Binary instrumental variables

• test_iv_conti.ipynb - Continuous instrumental variables

• test_fd.ipynb - Front-door adjustment

• test_jobs.ipynb - Real-world dataset example

These notebooks provide complete working examples you can run and modify.

Getting Help

If you get stuck:

• Check the ../user_guide/index for detailed explanations

• Look at the Standard CATE Estimation Example for complete code

• Read the API Reference for API documentation

• Open an issue on GitHub if you find a bug

Let’s Get Started!

Ready to begin? Start with Tutorial 1: Basics to learn the fundamentals!