Causal Inference [ESP48]

Detailed information about this course:

Description

Faculty: Prof. Miguel Hernán, MD & Dr. Sonja Swanson

The goal of many epidemiologic studies is to quantify the causal effect of a treatment (or exposure) on an outcome. In contrast, commonly used statistical methods provide measures of association that may lack a causal interpretation even when the investigator adjusts for all potential confounders in the analysis of a properly designed study.

To eliminate the discordance between the causal goals and the associational methods in epidemiology, it is necessary to a) formally define causal concepts such as causal effect and confounding, b) identify the conditions required to estimate causal effects, and c) use analytical methods that, under those conditions, provide estimates that can be endowed with a causal interpretation. These so-called g-methods can be used under less restrictive conditions than traditional statistical methods. For example, g-methods allow one to estimate the causal effect of a time-varying treatment in the presence of time-varying confounders that are affected by the treatment.

This course combines counterfactual theory and graph theory to present an integrated framework for causal inference from observational data, with a special emphasis on complex longitudinal data. The course presents the latest methodologic developments for the design and analysis of longitudinal studies. Specifically, the course will introduce the three g-methods (inverse probability weighting of marginal structural models; parametric g-formula; and g-estimation of structural nested models) in the setting of time-fixed treatments, and demonstrate inverse probability weighting for addressing causal questions regarding static and dynamic treatment strategies.

//Please note that the course information is subject to change and will be updated from time to time. We will do our utmost best to ensure the accuracy and reliability of the information on this website.//

Objectives

The student is able to:

• Recognize and formulate well defined questions concerning causal effects;
• Use causal diagrams to represent a priori subject-matter knowledge and assumptions;
• Identify the settings in which conventional methods for data analysis are inadequate;
• Use provided software to estimate causal effects using g-methods.
  

Participant profile

The course is intended for health researchers or other data scientists who will use observational studies to estimate causal effects as part of their current or future professional career. Examples include: epidemiologists, (bio-)statisticians, and other clinical or public health researchers.

Assessment

Attendance