# Extraction report: `turrietalndactionabilityjudgmentscause`

## Pipeline outputs used (in priority order)

- `papers/turrietalndactionabilityjudgmentscause/out/bundle.json` (index of outputs)
- Tables:
  - `papers/turrietalndactionabilityjudgmentscause/out/tables/tabula_stream_p5_t1.csv` (Experiment 1 descriptive stats + t/df/p/d)
  - `papers/turrietalndactionabilityjudgmentscause/out/tables/tabula_stream_p7_t2.csv` (Experiment 2 descriptive stats + t/df/p/d)
- Text/context:
  - `papers/turrietalndactionabilityjudgmentscause/out/fulltext.md`
  - `papers/turrietalndactionabilityjudgmentscause/out/tei.xml` (used for one moderator quote: “Christina observes …”; fulltext line was truncated)

Raw PDF was **not** consulted.

## Effect-size computations

Sampling variances `v` were computed reproducibly in:

- `papers/turrietalndactionabilityjudgmentscause/analysis/effect_sizes.qmd`
- Rendered audit record: `papers/turrietalndactionabilityjudgmentscause/analysis/effect_sizes.html`

Method used for both effects: `between_reported_d_t_df` (compute `v` from reported `d` + `t(df)` without needing split group Ns).

## Notes / potential issues

- Tables report `d` values as non-signed (positive) even when the mean difference is negative (Experiment 2 Knowledge row). In the extraction YAML, `d` is signed according to the project convention `d = mean(low) - mean(high)` using the reported group means.
- The per-condition split Ns are not reported in the extracted tables; group `n` is left as `null` in `groups[]`.

## Manual follow-ups (optional)

### HUMAN_CHECK_1 (optional)
Confirm bibliographic details for the citation string (journal name formatting, issue/pages) if you maintain a canonical citation style across the project.