נסקור נושאים בסיסיים ומתקדמים בהסקה סיבתית: 

אילו בעיות דורשות הסקה סיבתית? למה הסקה סיבתית קשה מחיזוי רגיל של supervised learning? כיצד ניתן ללמוד קשרים סיבתיים מנתונים? נלמד את הבסיס הרעיוני של התחום לפי Pearl ו-Rubin, כולל Causal graphs.

נביא דוגמאות מרפואה, כלכלה ומדיניות ציבורית, מדיה חברתית, שיווק ומכירות, ובריאות הציבור.

הקורס יכלול שלושה תרגילים להגשה ופרוייקט סוף קורס.

What problems require causal inference? Why is it harder than supervised learning? How can one learn causal relations from data? We will learn the approaches of Pearl and Rubin, including causal graphs.

We will bring examples from medicine, economics and public policy, social media, marketing and sales, and public health.

There will be three problem sets throughout the course, and a final project.

Syllabus (might change slightly):

1.    What problems are causal inference problems?
What is causal inference from observational studies? 
How is it related to supervised learning, and why is it harder?

2.    Conceptual foundations (I):
Rubin-Neyman potential outcomes
What are counterfactuals? Some philosophical background
What is causal identifiability?

3.    Practical methods (I):
Covariate adjustment, matching, propensity score

4.    Practical methods (II):
Covariate adjustment, matching, propensity score (cont.)
Case studies: Successes and failures of practical methods

5.    Conceptual foundations (II):
Graphical models, Pearl’s causal graphs, do-calculus, the backdoor criterion, Simpson’s Paradox; Practical methods through the lens of causal graphs

6.    Conceptual foundations (III):
Structural equation models
Identification by causal graphs, front door criterion

7.    Clever methods of identification:
Natural experiments, instrumental variables, regression discontinuity 

8.    Heterogeneous effects: theory and practice

9.    Causal discovery:
Logical and Bayesian methods, structure learning, Independence of Causal Mechanisms

10.  Connections to machine learning:
Covariate shift, contextual bandit problems, reinforcement learning

11.  Middle-of-project presentations

12.  Modern methods:
Causal forests, TARnet, CEVAE, deep instrumental variables, Causal Forests, X-Learner, Peysakhovich & Lada [and more]

Major References:

1.    Pearl, Causality

2.    Hernan, Miguel A., and James M. Robins. Causal inference. Boca Raton, FL:: CRC, 2010.
https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/

3.    Morgan & Winship, Counterfactuals and Causal Inference: Methods and Principles for Social Research (2nd edition, NOT 1st)

4.    Imbens, Guido W., and Donald B. Rubin. Causal inference in statistics, social, and biomedical sciences. Cambridge University Press, 2015.

5.    Peters, Elements of Causal Inference (http://www.math.ku.dk/~peters/elements.html)

6.    Pearl, Causal inference - an overview (http://ftp.cs.ucla.edu/pub/stat_ser/r350.pdf)

7.    Pearl, Glymour & Jewell, Causal Inference in Statistics: a Primer

8.    Angrist & Pischke, Mostly Harmless Econometrics

Rosenbaum, Observational Studies (2nd edition)