Introduction to Causal Inference

EPI 524 Subject & Catalog Number

Description

Causal inference is a core task in epidemiology. EPI 524 introduces a formal approach to causal inference based on a reasoning process that starts with explicit articulation of the causal research question. The ultimate choice of statistical analysis is then driven by this choice of question along with assumptions, including assumptions about confounding and confounders. This introductory course is limited to the setting where interest is in a causal effect of a time-fixed treatment on either a continuous or binary outcome.  

The material is generally split into two parts separated by two different types of causal assumptions. Part I will focus on reasoning under the assumption that all required confounders are measured in the data to be analyzed. Two types of statistical methods that may be justified for causal effect estimation under this reasoning will be introduced in Part I, standardization and inverse-probability weighting, where students will also learn to understand and reason critically about parametric model assumptions in the context of high dimensional confounding adjustment. Part II will focus on reasoning under alternative causal assumptions that do not require that all confounders are measured. In this case it is, instead, required that a so-called instrumental variable (IV) can be leveraged for causal inference. Statistical methods that follow from reasoning about IV conditions will be covered in this second part of the course.    

EPI524 is expected to be taken after EPI522.
 

This course is for MPH-EPI students only, no exceptions
Prerequisites: EPI522 and MPH-EPI only. May not be taken concurrently