Causal Inference in Practice 3-day Course at UCL

The short-course runs once a year, usually during the University reading week in February. Registration will open the Autumn before (October).

Causal inference is one of the most important and challenging aims in statistical and data science. Many fields, from clinical medicine to social science, seek to use empirical data to learn how different factors affect the world. This course covers the latest developments in causal inference methods and gives practical explanations about applying these methods to real research questions.

Course Content

• The causal roadmap and causal languages (potential outcomes framework)
• Directed acyclic graphs
• Target trial emulation
• Estimating marginal causal effects (Point Treatment) under “no unobserved confounding" using g-methods: g-computation, inverse probabiity of treatment weights and doubly robust estimation
• Instrumental Variable methods for obtaining causal effects under unobserved confouning"
• Introduction to Mediation analysis

Intended Learning Objectives

• Understand the potential outcomes framework and how to apply it
• Learn how to use directed acyclic graphs
• Estimate causal effects using observational data and the required assumptions
• Interpret causal analyses

What to expect?

The course is a mixture of lectures and applied practicals (approximately 1 to 1.5 hrs lecture and 1 to 1.5 hrs practice). The practicals include pen-and-paper exercises and data analysis in R or Stata. All of the data-based practicals will use simulated data and exemplar code that you can take with you.

Who Should Apply?

This course is aimed at people conducting applied epidemiological, medical, and other quantitative research, from PhD students to experienced researchers interested in learning more about causal inference methods. Participants should have familiarity with applied statistical analysis and an understanding of linear and logistic regression.

Lecturers

• Karla Diaz-Ordaz
• Neil Davies
• Bianca De Stavola