Registration for this module closes July 14. 

Live sessions will be held 8:30 a.m. - noon Pacific (11:30-3 p.m. Eastern) each day.

Mediation is a fundamental goal in many areas of research. Mediation methods seek to describe the pathways whereby a clinical treatment or public health intervention has an impact on downstream outcomes. Many methods have been developed over the years across many different literatures to tackle this problem, with varying degrees of statistical and causal assumptions. In this course, we will provide an overview of modern approaches to mediation analysis based on formal frameworks for casual inference. We will focus on precisely defining mediation effects and discussing assumptions needed to learn these effects from data generated in observational studies and clinical trials. Where possible, we will emphasize so-called multiply robust approaches that integrate modern machine learning methods to flexibly adjust for confounding while yielding valid statistical inference.

We will discuss at length methods for evaluating mediation of an intervention occurring at a single time point through a single mediator. We will also introduce the multiple time-point and multiple mediator extensions of these approaches. Methods will be illustrated using data from recent vaccine studies. Analyses will be illustrated in R but knowledge of R is not required for this module. In addition to lectures, the course will include in-class, hands-on activities to allow students to familiarize themselves with the methods and tools. The three-day course is geared towards health science researchers with at least basic experience in data analysis and statistics. A basic understanding of the following concepts would be helpful: confounding, probability (e.g., what is meant by the distribution of random variable, its mean and its variance), statistical inference (confidence intervals, hypothesis tests), and regression (linear and logistic). Advanced knowledge of these topics is useful, but not necessary. Equivalent UW SPH course pre-requisites are BIOS 511/512 (or BIOS 514/515). It is recommended, but not required that Module 6: Modern Statistical Learning for Observational Data be taken prior to this course.