An Unfriendly Introduction to Causal Inference

This book is short, it moves quickly, and it assumes you would rather learn the ideas than be eased into them.Most causal inference textbooks are written from within a single discipline: epidemiology, economics, or social science. They teach the subject through that field's conventions, notation, and running examples. If you come from physics, manufacturing, ecology, or nowhere in particular, you spend half your time translating before you reach the structural ideas that all of causal inference shares.This book presents those structural ideas directly. It follows a single thread — DAG → Test → Identify → Estimate — and every method in these pages is an elaboration of one of those four steps. The ladder of causation, the workflow from causal question to numerical answer, and the precise role of untestable assumptions are covered with as much mathematics as needed, in under 80 pages.What this book assumes: You are comfortable with probability, some linear algebra, and reading equations rather than skipping them. No prior knowledge of causal inference, graphical models, or potential outcomes is required.What this book does not do: Hold your hand.Throughout the text, pointers to R packages and their documentation connect each method to working code, a Python appendix will be available online soon. Read more