Overview
This page organizes my work on causal inference — from foundational concepts to applied modeling in marketing, experimentation, and decisioning.
The focus is on estimating causal effects, understanding heterogeneity, and making better decisions under uncertainty.
Foundations: From Association to Causation
Core concepts behind causal reasoning and why correlation is not enough.
- Association vs Causation: A Minimal Potential Outcomes Demo
- Causal Graphs, Confounding, Colliders, and Selection Bias
Randomized Experiments
Understanding causal effects through controlled experiments.
Observational Causal Inference
Estimating causal effects when randomization is not available.
- Propensity Scores in Practice: IPW and Doubly Robust Estimation
- Bank Marketing — Causal Linear Regression
Heterogeneous Treatment Effects (CATE & Uplift)
Understanding who responds, not just whether something works.
- Finding Who Actually Responds: CATE-Based Targeting
- Meta-Learners for Heterogeneous Treatment Effects
- Advanced CATE Estimation Methods
Bayesian & Decision-Focused Causal Modeling
Incorporating uncertainty into decision-making.
- Bayesian Models for Campaign Decisioning: Handling Uncertainty in Real Data
- Double Machine Learning Finds Segments, Bayesian Decides Which Ones to Trust
Structural Causal Models
Moving from treatment-effect estimation toward mechanism-level reasoning and counterfactual analysis.
Key ideas:
- Moving beyond average effect estimation
- Making assumptions explicit with causal graphs
- Using refutation and counterfactual reasoning
- Connecting estimation to intervention and policy analysis
How to Navigate
If you’re new to causal inference:
- Start with Association vs Causation
- Move to Experiments
- Then explore Observational methods
- Finally, go into CATE and Bayesian decisioning
Why This Matters
In real-world systems — marketing, product, and credit — decisions are not made on predictions alone.
They require understanding:
- What causes outcomes
- What works for whom
- How uncertain those effects are
This collection reflects that journey from theory → estimation → decisioning.