Archivierte Version! ⇒ Zur Aktuellen [de]…      Archived Version! ⇒ Go to Current [en]…
Statistik: mehr als Erbsen zählen

Sie sind hier:



Introduction to Causal Inference

Course description

A central goal of most scientific inquiries is to infer ‘cause and effect’ that describes what would have happened if the environment were changed through certain interventions. This course will provide an introduction to statistical methods for causal inference. We will cover causal inference under the assumption of no unmeasured confounder and under no such assumption. For the former, propensity score based methods will be the main technique and for the later, instrumental variable based methods. Some of the issues will be revisited based on graphical models that lead to clearer insights. If time permits, topics with matching, mediation, and interference will also be covered.

Potential topics include

  • Causal inference and potential outcomes
  • Randomized trials
  • Propensity scores for causal inference
  • Double robust estimation and semiparametric efficiency
  • Propensity score with high-dimensional covariates
  • Graphical models and the back-door criterion
  • Bias and confounding revisited from graphical models
  • Instrumental variables
  • Treatment effect heterogeneity with no-unmeasured confounders

Lecture and tutorial dates

Lecture:  Monday,  10:15 - 11:45,   Room: CDI 120

                Tuesday, 14:15 - 15:45,   Room: CDI 120

Tutorial:  Friday,     14:15 - 15:45,  Room: CDI 120

Note: The course will run from June, 17th to July, 29th. As an exception, the first lecture will take place during the tutorial slot on June, 17th from 14:15 - 15:45 in CDI 120.

The course material will be uploaded to the Moodle room for this course. Enrollment is possible using the key: Confounder

Prerequisites and eligibility

This course is taught in English and is targeted at Master students from Statistics, Econometrics and Data Science. It grants 4,5 ECTS towards the modules MS6/7, ME7 and MD-E1, respectively. Generally, an understanding of probabilistic and statistical concepts as well as linear algebra is strongly recommended. For Data Science students, successful completion of the Reading Courses is expected.

About the Instructor

Causal inference has always been a central theme for scientific discovery. The 2021 Nobel prize laureates in economics are two causal inference econometricians. The potential outcomes framework, which was first proposed by Jerzy Neyman in his 1923 Master's thesis and later extended by Donald Rubin, is fundamental to the counter-factual theories of causal inference whose advocates include the famous philosopher David Lewis. In 1943, Jerzy Neyman became the advisor of Erich Lehmann. In 1962, Lehmann became the advisor of Kjell Doksum. In 1980, Doksum became the advisor of Jeremy Taylor. In 2000, Taylor became the advisor of Menggang Yu.  

Textbooks (none required, but helpful for the course)

1.       Causal Inference for Statistics, Social, and Biomedical Sciences, Imbens and Rubin (Cambridge University Press)

2.       Causal Inference, Hernan & Robins (Available online)

3.       Observational Studies, Second Edition, Rosenbaum (Springer)

4.       Counterfactuals and Causal Inference: Methods and Principles for Social Research, Morgan & Winship (Cambridge University Press)

5.       Causality: Models, Reasoning and Inference, Second Edition, Pearl (Cambridge University Press)


For all questions regarding this course, please contact: