The Power of Why
A Visual Guide to Causal Inference
Moving beyond "what happened" to understand "why it happened." Causal inference is the science of identifying true cause-and-effect relationships from data.
The First Rule: Correlation is NOT Causation
Two trends can move together perfectly without one causing the other. This is the most common pitfall in data analysis.
Case Study: Ice Cream & Drownings
This chart shows a strong positive correlation. As ice cream sales rise, so do drowning incidents. Does ice cream cause drowning?
Of course not. The real cause is a...
HIDDEN CONFOUNDER
Hot weather! Higher temperatures cause people to buy more ice cream AND go swimming more often, leading to more drowning incidents. The heat is the true cause that links the two correlated events.
The Core Concepts
To think causally, we need a specific vocabulary.
Treatment
The intervention or variable whose effect we want to measure (e.g., a new drug, an ad campaign).
Counterfactual
The "what if" scenario. What would have happened without the treatment? This is the unobservable outcome we try to estimate.
Confounding
A third variable that influences both the treatment and the outcome, creating a misleading correlation.
The Ladder of Causation
A framework by Judea Pearl describing the three levels of causal reasoning.
1. Association (Seeing)
Observing how things are related.
"How are X and Y correlated?"
2. Intervention (Doing)
Predicting the effect of a deliberate action.
"What happens if I do X?"
3. Counterfactuals (Imagining)
Reasoning about "what if" scenarios.
"Why did X happen? What if I had acted differently?"
Methods at a Glance
How do we climb the ladder? Analysts use various methods to estimate causal effects.
Randomized Controlled Trial
Randomly assign groups to treatment or control. The "gold standard."
Difference-in-Differences
Compare trends over time between a group that got a treatment and one that didn't.
Regression Discontinuity
Exploit a cutoff rule to compare those just above and just below the threshold.
Instrumental Variables
Use a third variable (the "instrument") that affects treatment but not the outcome directly.
Why It Matters
Causal inference drives better decision-making everywhere.
Medicine
Does a new drug work? Are vaccines effective?
Policy
Does a job training program increase employment?
Business
Did our new feature actually cause user engagement to rise?