# Reproducible-analysis relay, 13:55–14:35

A small invented forage-drying study arrives with a draft report. Audit whether it runs and whether its scientific claims follow from the analysis. This practice question and dataset are separate from all four assessed projects.

## Supplied files and known design

`report-to-audit.qmd` is deliberately flawed. `draft-to-audit.md` supports auditing without software. `data/forage-drying.csv` contains **16 bags** in four forage-source blocks. Researchers group bags by four forage sources, then randomly assign two bags from each source to Shade drying and two to Warm drying. These source groups are called blocks. Each bag contributes one initial dry-matter mass and one retained dry-matter mass after the same drying period, both in grams. Three small laboratory portions were used to establish each bag's final mass estimate; the CSV gives one bag summary. Those portions improve the bag measurement; researchers do not assign a drying treatment separately to each portion. There are eight allocated bags per method, no missing summaries, and positive initial masses.

The planned comparison is Warm minus Shade in retained dry-matter fraction (`retained_dry_g / initial_dry_g`). The fraction ranges from zero to one; a difference can be expressed as percentage points by multiplying by 100. Conditions, duration, measurement calibration and generalisation beyond these source blocks require further documentation. All values are invented.

## Protect files

Make a working copy of the `relay` folder. Keep original source, data and log. Inspect before executing and work from the copied folder. Do not change input values to make prose match. The missing input-path problem is intentional; identify the supplied file before repairing the copied reference.

## 40-minute sequence

- **First 10 minutes:** attempt `quarto render report-to-audit.qmd --to html` from the copied relay folder, or inspect source if Quarto is unavailable. Record provenance, the question and first execution failure.
- **Next 10 minutes:** repair the path in the copy and inspect columns. Count bags assigned to each drying method; check outcome arithmetic, representation of source blocks, which group is subtracted from which and interval units. A successful render alone is insufficient.
- **Next 10 minutes:** trace each claim and label to a calculation. What diagnostic evidence is missing? Which source population can this study address? Prioritise consequential changes.
- **Final 10 minutes:** document a repair or specific proposed repair and rerun where possible. Compare original and revised conclusions; state what remains unverified.

## Audit log

| Claim/file location | Evidence/check | Problem or unresolved question | Proposed change | Rerun / inspected only |
|---|---|---|---|---|
| | | | | |

Include session information, relative paths and software limitations. Do not paste an untraceable result into a report. Apply the same reasoning to your own project afterward. If more than 40 minutes is needed, record the blocker and return during repair time.