Skip to main content

Causal Inference Course

Discussion 4. Exchangeability and Class Project

STSCI/INFO/ILRST 3900: Causal Inference

September 23, 2026

Exchangeability in DAGs

  • Exchangeability requires statistical independence \(A\perp\!\!\!\perp Y^a\)
  • Two nodes are dependent if and only if they are connected by an unblocked path (path with no colliders)
Exchangeability holds if all unblocked paths between \(A\) and \(Y\) are causal paths that point from \(A\) to \(Y\)

In-class Assignment 1

For each DAG determine whether exchangeability holds:

  1. List all paths between \(A\) to \(Y\)
  2. Cross out the blocked paths
  3. Exchangeability holds if all remaining paths are causal
Make sure to hand in the assignment before leaving!

Class Project

  • Divide into groups based on interests
  • Think of a causal question
  • Define the treatment and outcome variables
  • Find available data sources: ADD health or other (AI can be used for this part)
  • A detailed assignment will be posted (due toward the end of October)

In-class Assignment 2

To divide into groups please:

  1. Choose a topic of interest and fill out the first Google Form
  2. Find 3-5 students intersetd in same topic
  3. Fill out the second Google Form with names and NetIDs of group members