Causal graphs and d-separation
Write a causal story as a set of assignments and draw it as a directed graph. Three junctions, one rule for reading independence off the picture, backdoor paths, and the limits of what data can say about the arrows.
45 min
After this lesson you can
- Write a structural causal model and draw its graph
- Say what conditioning does in a chain, a fork and a collider, and why
- Apply the d-separation rules to decide whether two variables are independent given a set
- Identify backdoor paths and state the backdoor criterion
- State the faithfulness assumption and give a distribution that violates it
- Explain Markov equivalence and why data alone often cannot orient an edge
In this lesson
- Writing a causal story down
- Three junctions
- d-separation
- Backdoor paths
- Faithfulness: what the graph doesn’t promise
- Markov equivalence: the arrows data can’t see
This lesson is part of Descent Pro
Descent Pro unlocks every lesson on the route, from base camp to the summit. Every plan starts with a free trial.