Causal
Inference
(Biostatistics 140.665)
Meeting
times:
Instructor: Constantine E. Frangakis, Associate Professor, Biostatistics, Hygiene
E3642
cfrangak@jhsph.edu
Course description
An important task in public health and medicine is to evaluate and compare treatments, programs, and therapies. To make accurate evaluations, it is important to study (and respect) data on people, that is, which treatments we take and what outcomes we eventually have. For practical and ethical reasons, studies with people go beyond the experimental control found in fully laboratory settings, so people who take one treatment can generally be different prognostically from those who take another treatment. Causal inference means the framework for defining what we care about, for designing and analyzing studies, to take data we can observe between different treatment groups and correctly attribute them to effects of treatments. The course presents recent developments in designs and methods to better evaluate treatment effects.
The instructor acknowledges the sharing of ideas and material with Donald Rubin and Guido Imbens
Summaries of the lectures will be posted after each class
Lecture notes
Chapter
1. Introduction and framework
Chapter 2. Completely randomized assignment
Chapter 3. Treatment assignment with known and varying probabilities
Chapter 4. Ignorable treatment assignment and propensity scores Supplement on likelihood
Chapter 5. Studies with multiple treatments -- sequential ignorable assignment
Chapter 6. Studies with nonignorable noncompliance: instrumental variables
Chapter 7.
Studies with multiple partially controlled factors
Problem sets