A weeklong look at causal inference
Join us July 20–24 from 1 to 4 pm each day for this short intensive course, taught by Dr. Andrew Spieker, associate professor of biostatistics.
The course will be held in person at 2525 West End Avenue in Nashville, Tennessee.
Tuition for this course (in USD)
Regular: $950
Vanderbilt University / Vanderbilt Health faculty and staff: $700
Vanderbilt University / Vanderbilt Health students, trainees, and postdocs: $450
A 20% early bird discount will be applied to regular and employee registrations submitted between May 1 and May 31.
About this course
Many have likely heard that “correlation does not imply causation,” but that then begs the question: what exactly is causation in the first place? This five-day short course will provide a framework for modern causal inference. The first day will involve an overview of the potential outcomes framework and theory of DAGs (directed acyclic graphs). The second and third days will involve commonly implemented causal inference methods for use in cross-sectional data including standardization, matching, inverse-weighting, and instrumental variables. The fourth day will focus on methods for longitudinal data including marginal structural models and g-computation. The fifth day will likely feature miscellaneous advanced topics, which may include sensitivity analyses, parametric identification, and Bayesian methods. Throughout the course, emphasis will be placed on graphical representation of variables through DAGs and software-based implementation to real-world data.
Upon completion of this course, participants should be able to:
- explain the potential outcomes framework.
- explain causal identifiability assumptions.
- use directed acyclic graphs to characterize relationships between variables.
- choose between and implement causal methods suitable for real-world cross-sectional and longitudinal data.
- assess covariate balance and positivity violations.
Prerequisites
- A basic understanding of biostatistical methods, including linear and logistic regression.
- Prior experience working in R will be helpful, although it is not strictly necessary.
Details subject to change without notice