3 ms·
Even when you lack this level of control it may still be possible to infer causal effects. There's some interesting work that I've seen on this, good write up h
by n4r9 7y ago
Even when you lack this level of control it may still be possible to infer causal effects. There's some interesting work that I've seen on this, good write up here: http://www.michaelnielsen.org/ddi/if-correlation-doesnt-imply-causation-then-what-does/ http://www.michaelnielsen.org/ddi/if-correlation-doesnt-impl...
- raindeer3 7y agoCan recommend Judea Pearl's recent book on the subject. The Book of Why
- dhshahsndeisjwn 7y agoWas just about the say the same. He has an excellent chapter on the history of smoking research with an interesting anecdote (repeated here https://priceonomics.com/why-the-father-of-modern-statistics-didnt-believe/ https://priceonomics.com/why-the-father-of-modern-statistics...) that the pipe-smoking Fischer refused to believe smoking caused harm well after the evidence was clear because correlation!=causation.