Summary
The episode works through four questions that students commonly find difficult in the study of measures of association. It uses a hypothetical cohort resampled as a case-control study to show why the odds ratio is the ratio measure a case-control design supplies directly, and it draws on a 1999 study of referrals for cardiac catheterization to illustrate how an odds ratio overstates an association when the outcome is common and the odds ratio is read as a risk ratio. It then compares numbers needed to treat across groups with different baseline risks, and it uses Health Canada's estimates of lifetime lung cancer risk from radon and smoking to contrast the ratio and difference scales and to explain why attributable fractions for different causes can sum to more than one hundred percent. The hosts close by recovering a point estimate and a p-value from a confidence interval and by discussing whether the label of statistical significance should be retired, and they agree on a shared rule for reporting estimates, intervals and exact p-values.