Hacker News new | ask | show | jobs
by astrange 1000 days ago
You can't demonstrate causation by controlling for things, because of collider bias.

https://en.wikipedia.org/wiki/Bad_control

You need better methods like an intervention study or a natural experiment to show causation.