# Extraction report: shurakovndstakeseffectnew

## Summary
- Pipeline outputs used (preferred sources):
  - Extraction index/status: `papers/shurakovndstakeseffectnew/out/bundle.json`
  - Study methods + coding details + retraction rates: `papers/shurakovndstakeseffectnew/out/fulltext.md`
  - Table-based numeric values (means/SDs and reported d): `papers/shurakovndstakeseffectnew/out/tables/tabula_stream_p12_t2.csv`
  - Additional table context (binary χ² tests): `papers/shurakovndstakeseffectnew/out/tables/tabula_stream_p10_t1.csv`
  - Paper metadata cross-check (year/DOI): `papers/shurakovndstakeseffectnew/out/tei.xml` and rendered page image `papers/shurakovndstakeseffectnew/out/images/pages/p001.png`
- Output YAML: `papers/shurakovndstakeseffectnew/shurakovndstakeseffectnew.yaml`
- Effect sizes:
  - Sampling variances `v` for `s1_e1`, `s2_e1`, `s3_e1` computed in `papers/shurakovndstakeseffectnew/analysis/effect_sizes.qmd` from reported `d` + n_low/n_high=100.

## Manual interventions / non-pipeline sources (required disclosure)
- External dataset used (public OSF project linked in the paper):
  - Source: https://osf.io/tys3p
  - Downloaded into `papers/shurakovndstakeseffectnew/data/`:
    - `papers/shurakovndstakeseffectnew/data/Experiment_1_analized_responses.csv`
    - `papers/shurakovndstakeseffectnew/data/Experiment_1_all_responses.csv`
    - `papers/shurakovndstakeseffectnew/data/Experiment_2_analyzed_responses.csv`
    - `papers/shurakovndstakeseffectnew/data/Experiment_2_all_responses.csv`
  - Used to reproduce the Table 2 descriptives (means/SDs) and sanity-check the reported Cohen’s d values (see `analysis/effect_sizes.qmd` output).

## Notes / potential limitations
- Experiments were extracted at the “subcondition” granularity recommended by `docs/extraction_instructions.md`:
  - Separate `studies[]` entries for Experiment 1 first-person vs third-person, plus Experiment 2 (modified design).
- Moderator coding update: after clarifying that `awareness`, `evidence`, and `evidence_reliability` are coded from the subject
  of the target knowledge attribution, third-person Experiment 1 effects `s2_e1` and `s5_e1` are coded `awareness: No`.
  Peter's evidence remains `First Person` because his basis is his own remembered prior bank visit.
- Reported p-values in Table 2 are thresholds (“< .001”), so `reported_test.p` is set to `null` and the threshold is preserved in `reported_test.notes`.
- OSF data check vs Table 2:
  - Means and SDs for Neutral/Stakes match Table 2 when using the “first 100 valid responses per condition” rule described in the paper.
  - For Experiment 1 (first-person), the Table 2 reported `d=1.06` does not match `d≈1.17` when computed from those reproduced means/SDs using a standard pooled-SD Cohen’s d formula. This is flagged with `effect_size.needs_review: true` for `s1_e1`.

## Open questions (human input requested)
### Q1 — Which d should be used for Experiment 1 (first-person) Stakes vs Neutral?
- Current YAML uses the paper’s Table 2 reported `d=1.06` (with `needs_review: true`).
- OSF data check suggests `d≈1.17` (pooled-SD) for the same comparison (means/SDs reproduce cleanly).
- If you want, I can update the extraction to use a computed-from-groups `d` (from OSF data or from Table 2 means/SDs) for `s1_e1` for consistency.

## Validation / rendering
- YAML schema validation: `papers/shurakovndstakeseffectnew/shurakovndstakeseffectnew.yaml` validates against `docs/stakes_meta_schema.json` (via `jsonschema`).
- Quarto computation record: `papers/shurakovndstakeseffectnew/analysis/effect_sizes.html` rendered from `papers/shurakovndstakeseffectnew/analysis/effect_sizes.qmd`.
- HTML extraction report: `papers/shurakovndstakeseffectnew/shurakovndstakeseffectnew.html` generated via `invoke render-extraction shurakovndstakeseffectnew`.