8.9 Using paths to test the DAG

  1. List all of the paths between A and B.
  2. Consider only the open paths.
  3. Do what you need to do to make sure they’re all Closed.
  4. If A and B are still related to each other (placebo test), that means there must be some other path you didn’t account for.

Interpretation of placebo tests:

  • We have a model, so not accounting for all effects is okay.
  • small but nonzero relationship: fine
  • enormous and super-strong relationship: 😱