# Extraction report: pinillos2012knowledgeexperimentspracitcal

## What I used (pipeline outputs)

- `papers/pinillos2012knowledgeexperimentspracitcal/out/bundle.json`
- `papers/pinillos2012knowledgeexperimentspracitcal/out/text.txt`
- `papers/pinillos2012knowledgeexperimentspracitcal/out/fulltext.md`
- `papers/pinillos2012knowledgeexperimentspracitcal/pdf/paper.pdf`

## Tables / figures

- `out/tables/` is empty (Camelot/Tabula extractions were filtered out).
- Numeric evidence was taken from the text/footnotes and validated against PDF page images.

## Computations (reproducible)

- Source: `papers/pinillos2012knowledgeexperimentspracitcal/analysis/effect_sizes.qmd`
- Rendered audit: `papers/pinillos2012knowledgeexperimentspracitcal/analysis/effect_sizes.html`

Methods used:

- Between-subject Mann-Whitney `r -> d`: `esc::esc_rpb(..., es.type = "d")`
- Within-subject paired contrast: `metafor::escalc(measure = "SMCC", ti = t, ni = n_total)`

## Assumptions applied

- For contrasts where split group Ns were reported, those Ns were used directly in `esc::esc_rpb`.
- For contrasts where only total N was reported, equal split was assumed (`n_low = n_high = totaln/2`) and `esc::esc_rpb(..., totaln = N)` was used.
- For the Belief-fixed probe, reported `r=.34` appears unsigned while `Z=-2.98`; sign was set by direction coding (`d = mean(low) - mean(high)`) using the reported medians.

## Current status

All coded effects in YAML now have numeric `d` and `v` and are linked to the audit QMD.

Study 3 remains a special case for downstream synthesis:

- The within-subject juxtaposed effect is coded with `SMCC` because the paper reports a paired `t` on transformed differences, but not the condition-level SDs needed to recover an `SMCRP` effect.
- The YAML now flags this effect as a metric-mismatch sensitivity case for mixed-design meta-analysis.