Causal inference is the problem of estimating the effect of an intervention given a causal model and observational data. It differs from causal discovery in that the causal model is given; it differs from probabilistic inference in that it asks for an interventional probability—and is given a causal model---instead of asking for a conditional probability.
Note: causality is semantically overcharged. Here, it refers to the information extracted from causal interventions.
Notes.
- back-door adjustment formula
- front-door adjustment formula
- solving causality problems via conditionals
- non-solvable identifiability problems
- smoking causes cancer problem
- Napkin problem
References.
- Causality (Pearl, 2009)
- Causal Inference by String Diagram Surgery (Jacobs, Kissinger, Zanasi)
- Basic Causal Inference via String Diagrams (Piedeleu, 2023)
- Identification of Conditional Interventional Distributions (Shpitser and Pearl, 2006)
- A Crash Course in Good and Bad Controls (Cinelly, Forney, Pearl, 2022)