# Extraction report: phelan2014evidencestakesmatter

## What I used (in required priority order)
- `papers/phelan2014evidencestakesmatter/out/bundle.json` (index; tables extraction produced no usable tables)
- `papers/phelan2014evidencestakesmatter/out/fulltext.md` (all numeric results and quotes)
- `papers/phelan2014evidencestakesmatter/out/tei.xml` (DOI)
- I did **not** consult `papers/phelan2014evidencestakesmatter/pdf/paper.pdf` directly; all `provenance.page` are therefore `null`.

## Computations performed (reproducible)
- `papers/phelan2014evidencestakesmatter/analysis/effect_sizes.qmd` computes all `effect_size.d`/`effect_size.v` used in `papers/phelan2014evidencestakesmatter/phelan2014evidencestakesmatter.yaml`.
  - Between-subjects: `v` computed from reported `d` + `t(df)` without split group Ns (`between_reported_d_t_df`).
  - Within-subjects (Two Cases): `d` computed from means/SDs (SMCRP) and `v` computed using a within-person `r` recovered from paired `t(df)` (`within_smcrp_t`).

## Notes / decisions
- `out/tables/` was empty (Camelot/Tabula extractions were filtered out), so all statistics were taken from text/notes in `out/fulltext.md`.
- The paper contains a “Fourth Study” (general commitment test: % selecting an “importance” factor). It was **not** encoded in the extraction YAML because the schema requires each `studies[]` entry to contain at least one high-vs-low-stakes effect (`effects[].contrast`), which the commitment test does not provide.
- The paper reports some `d` values as unsigned; for consistency the extraction follows `contrast.sign_convention: d = mean(low) - mean(high)` (e.g., Study 3 non-juxtaposed yields a negative `d` because the Important mean exceeds the Unimportant mean).

## Uncertainties / follow-ups (human input fields)

### HUMAN_CHECK__CITATION_DETAILS
Please confirm whether you want `paper.citation` to include journal volume/issue/pages (currently minimal + DOI only).

### HUMAN_CHECK__INCLUDE_SUPERSEDED_DRUNKS_STUDY
Study 4 in `papers/phelan2014evidencestakesmatter/phelan2014evidencestakesmatter.yaml` is an earlier “two drunks” version reported only in notes and described as superseded. Confirm whether you want it kept (default: kept, flagged in labels/notes) or excluded from meta-analysis inputs.