The class, for the most part, builds on the following three references, which—in the course schedule below—will be abbreviated as ELW, HR and PGJ, respectively:
- Elwert, F. (2013). Graphical causal models. In: Handbook of Causal Analysis for Social Research. Ed. by Morgan, S. L. New York: Springer, pp. 245–272. = ELW
- Hernán, M. A. & Robins, J. M. (2018). Causal Inference Boca Raton, FL: Chapman & Hall/CRC, forthcoming. = HR.
- Pearl, J., Glymour, M., & Jewell, N. P. (2016). Causal Inference in Statistics: A Primer. West Sussex, UK: Wiley = PGJ.
Introduction (Oct 8)
Key point: Causal inference is important!
Further reading: Petersen & van der Laan (2014)
Assignment 1 (due Oct 10, 5:44pm)
Data and correlations (Oct 10 & 15)
Key point: Correlations describe group differences in the data (whereas causal effects capture changes in the world)!
Required reading: PGJ, Preface + HR, Preface to Ch. 1 + Shalizi (2018), Ch. 1.1
Further reading: PGJ, Ch. 1.3 (before 1.3.10) + Shalizi (2018), Ch. 1.2
Assignment 2 (due Oct 15, 9:59am)
Modelling correlations (Oct 17 & 22)
Key point: Statistical models (like regression) interpolate and extrapolate correlations in sparse data (but don’t estimate causal effects)!
Required reading: HR, Ch. 11
Further reading: PGJ, Ch. 1.3.10 + 1.3.11 + Shalizi (2018), Ch. 1.3-1.5 + 11
Assignment 3 (due Oct 22, 9:59am)
The limits of correlations (Oct 24 & 29)
Key point: Data, correlations, and statistical models alone are not sufficient for causal inference (we also need causal models)!
Required reading: Shalizi (2018), Ch. 2.4 + PGJ, pp. 1-6
Further reading: Shalizi (2018), Ch. 2
Assignment 4 (due Oct 29, 9:59am)
Building causal models (Oct 31 & Nov 5)
Key point: Causal models are theories/assumptions about the common causes of our phenomena of interest!
Required reading: ELW, pp. 245-249
Further reading: PGJ Ch. 1.4 + 1.5
Assignment 5 (due Nov 5, 9:59am)
From causal models to correlations (Nov 7 & 12)
Key point: Causal models imply a set of correlations we should see in data (but the same set of correlations is consistent with multiple causal models)!
Required reading: ELW, pp. 249-254
Further reading: PGJ Ch. 2
Assignment 6 (due Nov 12, 9:59am)
Defining and identifying total effects (Nov 14 & 19)
Key point: Causal inference addresses what-if questions by separating causal correlations from noncausal correlations on the basis of causal models (which is often very tricky and sometimes impossible)!
Required reading: HR Ch. 1 + 3 + 6.4 + 4.1 + 4.3 + 6.6
Further reading: ELW, pp. 254-261 + PGJ Ch. 3 + 4
Assignment 7 (due Nov 19, 9:59am)
Confounding bias (Nov 21 & 26)
Key point: Failing to control for specific variables can bias causal inference!
Required reading: HR Ch. 7 (except 7.4)
Further reading: ELW, pp. 261-262
Assignment 8 (due Nov 26, 9:59am)
Collider bias and overcontrol bias (Nov 28 & Dec 3)
Key point: Controlling for specific variables can bias causal inference, too!
Required reading: HR Ch. 8
Further reading: Elwert & Winship (2013) + Hernán et al (2004)
Assignment 9 (due Dec 3, 9:59am)
Randomized experiments and bias (Dec 5 & 10)
Key point: Randomization is a design-based solution for some (but not all) potential biases in causal inference!
Required reading: HR Ch. 2.1 + PGJ Ch. 4.3.3
Further reading: Deaton & Cartwright (2018) + Mansournia et al (2017) + Sampson (2010)
Assignment 10 (due Dec 10, 9:59am)
Adjusting for covariates using regression (Dec 12 & 17)
Key point: Under specific conditions, regression can be useful for causal inference!
Required reading: HR, Ch. 10.1 + 10.2 + 10.5 + 15.1.
Assignment 11 (due Dec 17, 9:59am)
Adjusting for covariates through re-weighting (Dec 19 & Jan 7)
Key point: Regression is one of many statistical tools that can be used for causal inference!
Required reading: HR, Ch. 2.4 + 12 (except 12.3 + 12.6).
Assignment 12 (due Jan 7, 9:59am)
Exam due (Jan 30)
Key point: Register on time for the exam on Klips!