# Beyond the Published Portfolio · A11 methods guide

Corresponding original protocol for the volume's conceptual article **Publication as Learning Infrastructure**. Author: Dr. Freedom Cheteni. Official publisher: The Research University. Article version 1.0, prepared 15 October 2026 Pacific; source retrieval/verification 16 October 2026 UTC.

## What has actually been completed

- A frozen DataCite REST metadata audit: 300 unique DOI records, three 100-record resource-type strata, education subject query, publicationYear 2024–2025, first page sorted by DOI name.
- Nine artificial-data paired-mentor design scenarios, 20,000 replicates each,seed 20261016. Power, null rejection, coverage and Monte Carlo error recorded.
- Offline executable tools and tests. No participant outcomes, ethics/registration approval, actual study allocation, ORCID authentication/writes, DOI registration or acceptance decision exists in this package.

## What the prospective study proposes

96 mentor-course clusters,48 pairs,12 baseline learners/cluster, target 1,152. Twenty weeks comparing an optional supported publication pathway with time/resource-matched private authentic-audience work. Primary masked transfer change 0–24; gated provenance mastery 0–16. Publication voluntary, independent editorial decisions, no grade/follow-up penalty for keeping work private. Exact design/analysis/instruments are in their files; these require independent implementation and ethics review before recruitment.

## Reproduce completed analyses

Python 3.12.14, NumPy 2.3.5, SciPy 1.17.0 were used. Create your own isolated environment and install versions in requirements.txt. Standard-library metadata audit does not need NumPy/SciPy.

```sh
python research_tools.py audit --input metadata-projection.json --output reproduced-metadata-audit.json
python research_tools.py simulate --output reproduced-design-analysis.json
python tests.py
```

Simulation results are artificial, not observed educational effects. Pair-diagnostic accepts a JSON object containing a 1D numeric `pairDifferences` list and produces a complete-pair mean/t interval and Fisher sharp-null sign-flip check. It does not implement the primary missing-data model or validate trial conduct. Never use it to silently delete incomplete clusters.

## Frozen DataCite records and rights

frozen-query.json records exact three URLs, source-byte SHA-256 hashes, byte counts and UTC retrieval times. Source responses contain 100 Dataset, 100 Software, 100 Dissertation records. The bounded query is not representative, random or exhaustive; unseen pages were not inspected. The DataCite API's total/totalPages values are retained without interpreting them as complete counts inspected. Live queries may return changed metadata.

metadata-projection.json is the frozen audit input, minimized to audit fields. It omits creator names, actual creator ORCID values and abstracts. It retains publisher/rights/related-resource metadata needed for the audit. source-audit.json records its SHA-256. Original full responses are retained outside public materials; their source hashes are supplied. The projection reproduces the published counts, not every omitted source field.

DataCite's policy applies CC0 to deposited DOI metadata available through its APIs: https://support.datacite.org/docs/datacite-data-file-use-policy . The waiver does not license linked resources or remove individual privacy/publicity rights. Attribution: DataCite and the depositing repositories. No linked dissertation, dataset, code file or source full text is redistributed. No resource is classified ingestion-ready from these metadata.

## Interpretation boundaries

- Creator ORCID metadata/checksum validity does not authenticate a record holder, verify authorship or identify a student.
- DOI presence/resource type does not prove peer review, consent, academic quality or learning.
- Resource version, metadataVersion and typed version relations differ.
- publicationYear, Issued, registry-created, registered and top-levelpublished timestamps differ. A null published timestamp does not mean unpublished; a null contentUrl does not mean inaccessible.
- No new primary-study effect is reported. Conceptual-companion review does not transfer to this protocol. References mark document/full-text vs abstract/metadata availability individually.

## Files

article.md;study-specification.json;instruments.md;analysis-specification.md;research_tools.py;tests.py;requirements.txt;metadata-projection.json;metadata-audit.json;design-analysis.json;source-audit.json;frozen-query.json. Root publication workflow adds the final standalonePDF and ZIP. No secrets or participant records belong in this package.

The API schemaVersion is retained as supplied (kernel-4); it does not certify compliance with every property of the latest 4.7 documentation. Resource metadata, identity authentication and editorial decisions remain separate.