The directed acyclic graph (DAG) has become widely recognized as a useful tool for contending with confounding and selection bias in study design and data analysis. This workshop will introduce participants to the basic concepts, terminology and rules for drawing DAGs and interpreting them. We will focus on their use in identifying appropriate and inappropriate sets of covariates for restriction, matching, stratification and adjustment. As we work through concrete examples of increasing complexity, we will become adept at user-friendly, free software for analyzing DAGs. Additional topics will include the use of DAGs in estimating direct and indirect effects in mediation analysis and in assessing bias from missing data and measurement error. Throughout, we will compare and contrast the use of DAGs with traditional concepts and tools. To get the most from the workshop, participants should bring a computer with the slides, examples and software.
Workshop Instructor: Charles Poole