Causal Inference
Learn what causal inference means, why assumptions matter, and how to estimate treatment effects step by step.
Background reading
Introduction to Causal Inference
Causal inference is the process of learning how an outcome would change if we intervened and changed a treatment or action. The important comparison is what would happen under one action versus what would happen under another.
Example: A retailer sends a promotion and the customer spends 30 dollars. Did the promotion cause all 30 dollars of that spending? No: the customer might have spent 25 dollars without it. In that hypothetical case, the promotion caused an increase of 5 dollars. The challenge is that we do not observe both outcomes for the same customer on the same occasion.
We will first name the two outcomes we wish we could observe, then ask when another customer can provide a fair comparison. Only after that will we estimate effects with stratification and weighting.
The difficult part is not fitting a model. It is deciding which comparison can answer the causal question and stating the assumptions that make that comparison credible.
Loading course…