Data Science
Advanced
Causal Inference for Data Science
Move beyond correlation to measure true causal impact in observational business data.
Course Overview
Implement propensity score matching, instrumental variables, and causal graphs using Python DoWhy to evaluate product interventions without A/B testing.
What You Will Learn
Formulate causal questions using Directed Acyclic Graphs (DAGs) and Structural Causal Models
Identify confounding, collider bias, and selection bias in observational data
Implement Propensity Score Matching (PSM) and Inverse Probability Weighting (IPW)
Apply Difference-in-Differences (DiD) and Synthetic Control methods
Estimate heterogeneous treatment effects using Causal Forests and DoWhy
Tools & Technologies Used
DoWhy
CausalML
Statsmodels
Python 3.11
Scikit-Learn
Structured Curriculum
2 Modules · 4 Lessons · 220 Minutes Total
Module 1
Module 1: Causal Graphs & Confounding Identification
2 lessons
Module 1: Causal Graphs & Confounding Identification
Draw causal DAGs and isolate treatment effect mechanisms.
-
Structural Causal Models & DAG Design
Identify back-door paths, front-door paths, and colliders in causal graphs.
-
Propensity Score Matching & Weighting
Balance treatment and control groups using logistic propensity scoring.
Module 2
Module 2: Quasi-Experiments & Modern Causal ML
2 lessons
Module 2: Quasi-Experiments & Modern Causal ML
Apply DiD, Synthetic Controls, and Machine Learning Causal Forests.
-
Difference-in-Differences & Synthetic Controls
Estimate treatment effects across parallel historical trends.
-
Heterogeneous Treatment Effects with DoWhy
Use Microsoft DoWhy to refactor observational data analysis into 4 step causal checks.
Practical Project & Capstone Outcome
🚀 Capstone Project
Product Pricing Change Causal Impact Analysis
Analyze an observational dataset of product feature adoption to measure true revenue lift while controlling for marketing spend and user demographics.
Prerequisites
- Python intermediate proficiency
- Basic understanding of probability and regression
Intended Audience
- Data Scientists evaluating product features without randomized control trials
- Quantitative Researchers making policy or pricing recommendations
Instructor Information
D
Dr. Lars Mikkelsen
Course Author & Industry Expert
Dr. Lars Mikkelsen is a Principal Econometrician with 15+ years of experience applying causal inference to technology products.
Frequently Asked Questions
Why is causal inference necessary if we already do A/B testing?
A/B tests are expensive or unethical for many decisions; causal inference unlocks insights from historical data.