Summary
The episode works through four problems that graduate students commonly meet when they turn qualitative codes into counts. It shows how the choice of recording unit can reverse a comparison, why passages from one participant break the independence that a significance test assumes, and how a percentage from a purposive interview corpus differs from a percentage that describes a defined body of texts, using a Canadian study of newspaper coverage of the opioid crisis. Sarah drafts a codebook entry for an invented latent code, discovers that her inclusion rule works as a word list, and revises it with exclusion rules, typical and atypical examples and a paragraph-level context unit. The hosts then show how agreement checks built from each coder's own highlights leave out the passages only one coder marked and hide disagreements about unit boundaries, work through an invented newspaper corpus in which a framing trend reverses once the articles are split by outlet, and debate whether small purposive studies should report p-values before agreeing on a shared reporting rule.