Discussion questions

1. Bayes factors evaluate relative evidence between models, not just whether one is “rejected”. If inference becomes a model comparison problem, how does that change how we design studies? Do we need to think more carefully about what 𝐻1 actually is (instead of treating it as a vague “not H0”)? Does this force better theory or just more complicated modeling?

2. The lecture argues that p-values do not quantify evidence, while Bayes factors explicitly compare evidence between hypotheses. If researchers have been using p-values as if they measure evidence for decades, what are the practical consequences of this mismatch? Are entire literatures subtly misinterpreted or is the distinction more philosophical than impactful?