Read on Twitter
Broadly speaking I think all schools of thought on causal inference are the same. People love to argue weird nuances and fight about DAGs, but everyone is more or less on the same page. In my experience the the biggest differences come from the foundation: 1/
Causal Inf stems from either:
1. What is your actual estimand of interest?
or
2. If there were no ethical or financial barriers, could you run an RCT to answer your question?
2/
1. What is your actual estimand of interest?
or
2. If there were no ethical or financial barriers, could you run an RCT to answer your question?
2/
So if we were to ask "does smoking cause cancer?" we can fit into both frameworks by adding some detail/precision:
1. What is the added increase (if any) in lifetime cancer chances that comes from smoking 1 pack a day from age 18 to 60?
3/
1. What is the added increase (if any) in lifetime cancer chances that comes from smoking 1 pack a day from age 18 to 60?
3/
2. If I randomly assigned 1000 people to smoke one pack a day from ages 18 to 60 and 1000 people to never smoke, what would be the difference in total cancer incidence?
Similar ideas, right? It gets weirder when questions get harder
4/
Similar ideas, right? It gets weirder when questions get harder
4/
Say I want to understand the "treatment effect" of race on probability of getting cancer? Or on the chances of getting an interview? I can maybe get at this in framework 1, but I can't randomly assign race in framework 2.
5/
5/
Anyway. If you are intrigued by causal inference, here's a previous thread of mine on some recommendations for books/papers/sites etc. https://twitter.com/CausalKathy/status/1329807280498466819
