The objective of this course is to train students in statistical methods that allow for the estimation of causal relationships, using randomized experiments or quasi-experiments. These methods involve either implementing controlled experimental protocols or leveraging statistical data to exploit “natural” experiments or social, economic, or institutional events, which under certain assumptions, produce differentiated exposure to treatment among various populations, making a causal interpretation plausible.

The main chapters are:

  1. Rubin Causal models and RCTs
  2. Imperfect compliance
  3. Instrumental variables and LATE
  4. Difference-in-difference
  5. Regression discontinuity
  6. Design-based inference
  7. Experimental designs
  8. Introduction to Machine learning for causal models
  9. Topics

ECTS: 6

Grading will be based on quizz routinely filled before or during lectures and a final exam.

Prerequisites: basic notions of probability (such as conditional expectation) and regression; be comfortable with elementary maths. Test theory will not be covered in the course, but students are expected to be familiar with a Student t-test in a regression framework. Students from Cognitive Science will have a specific hands-on course to practice R, manipulate data, revise regression and testing, and apply methods. Students from the Economics department will have to prepare an econometrics project.