Certain people have a better way of thinking about things. When they’re in a lecture, they ask better questions. When they’re speaking, their train of thought is clearer. These are the people, like Gilbert Strang, Matt Devine, Bartoz Milewski, Mathew Johnson, that you want to learn from.
I am certainly not one of these people. I am firmly in the land of the confused. The land of the don’t know much, write bad code, begin paragraphs with “and another thing I’ve thought of”. The one thing that I believe I share with the people above, though, is that I enjoy thinking about how to think about things.
The following is a therefore a collection of notes that walks one through how I think about causal inference.
Regular Conditional Probability
There are roughly four identification strategies used in causal inference. I would argue that the common identification strategies in the economics literature not mentioned below - like regression discontinuity design, synthnetic controls, etc — are in fact special cases of those mentioned below.