Summary
The episode works through four questions that students often find difficult in the introduction to causal concepts. It revisits John Snow's 1854 comparison of the Southwark and Vauxhall and Lambeth water companies as an early argument for exchangeability, and it uses a hypothetical trial to show that an average treatment effect describes a net effect across a group and cannot identify which individuals were helped or harmed. It then uses a made-up clinic example and the birth weight paradox to show how adjusting for a collider can create a spurious association. A harder hypothetical example shows why the attributable fractions for two components of the same sufficient cause can sum to more than one hundred percent, and how the risk ratio for one component changes with the prevalence of its complement. The hosts close by debating the Bradford Hill viewpoints and agree that temporality is the one firm requirement and that the other viewpoints serve as prompts to look for chance, bias and confounding.