Causal Inference: Why We Should and How We Can Teach it in Introductory Courses

Abstract

We are living in a world full of multivariate observational data. Qualitative assumptions about the data generating process, operationalised in simple directed acyclic graph can help students to understand multivariate phenomena like Simpson's or Berkson's paradox, confounding and bias. By teaching causal inference the introductory course can overcome the mantra "correlation does not imply causation".

The webinar discusses some motivation as well as teaching ideas and the integration in the curriculum.