Summary
The episode works through how an analyst can tell a theme from a pattern produced by the interview guide, by an overly broad code, or by a word concentrated in one interview, and why a theme is stated as a claim about shared meaning. The hosts code an invented excerpt line by line and show why one grouping of codes is a category while another, which draws on all four metacoding bins, is a candidate theme. Sarah drafts a codebook entry aloud, finds that her inclusion criteria leave out the link to loneliness that her brief definition contains, and revises the entry with a negative example, a rule for a neighbouring code and a memo. A harder worked example shows how a rare code produces high percent agreement with a low kappa, and how a perfect kappa can rest on a single passage. The hosts then debate whether intercoder reliability belongs in interpretive coding and agree to reserve agreement statistics for coding-reliability work, reported code by code with how often each code was applied.