Replication Package: Trust Infrastructure for Global Research-Funding Data - Building, Auditing, and Hardening a Multi-Source Evidence Array

Anonymous replication deposit accompanying the double-blind submission of the study Trust Infrastructure for Global Research-Funding Data: Building, Auditing, and Hardening a Multi-Source Evidence Array. This deposit contains the frozen, author-derived tables and the audit statistics sufficient to reproduce every quantitative claim in the study that can be redistributed under the license terms of the underlying sources. It contains no source snapshots, no raw extraction dumps, no manuscript files, and no figures. Deposit date: 2026-10-30. Frozen data baseline: 2026-10-14 (manifest) / 2026-10-21 (refresh audit). Contents (inside the reassembled ZIP): data/ - six frozen terminal edge and reference tables in Parquet format plus three frozen phase-statistics JSON files (audit trail); scripts/ - reproduce_study.py, which recomputes 67 reported quantities from the frozen tables and compares each against the study's reported value; docs/ - the five-entry discrepancy ledger (D1-D5) and a summary of the 34-source frozen manifest with the 2026-10-21 refresh audit. Reassembly. The archive is stored as split volumes of one ZIP file. Concatenate all files named ReplicationPackage_v1.zip.NN.part in ascending numeric order into a single file ReplicationPackage_v1.zip (Unix/macOS: cat ReplicationPackage_v1.zip.*.part > ReplicationPackage_v1.zip; Windows cmd: copy /b ReplicationPackage_v1.zip.*.part ReplicationPackage_v1.zip). The expected SHA-256 of the reassembled file is recorded in ASSEMBLY.txt inside this deposit. After reassembly, run scripts/reproduce_study.py (Python 3.10+, pandas, pyarrow) from the scripts/ directory to reproduce all 67 quantities. Items not redistributed, with reasons: (1) Europe PMC acknowledgement-extraction population - third-party terms permit sharing counts, rule sets, and samples, not full extracts; frozen counts are in the phase-statistics JSON files. (2) NIH RePORTER public registry extract - public-domain agency product obtainable directly from the provider; exact query parameters and date stamps are documented in the source-manifest summary. (3) Work-title fields used by the DOI collision scan - input counts are reproducible; the scan outcome is recorded in the discrepancy ledger. (4) Internal corpus-to-work crosswalk (ledger entry D4) - internal mapping table. (5) Unregistered in-progress source snapshots - recorded as growth during the freeze window in the manifest summary. Versioning and provenance. All tables are the frozen terminal artifacts of the audit window (2026-10-14/2026-10-21). Downstream corrections were issued as versioned re-analysis releases (v2), as described in the study. Code: MIT. Data: CC-BY-4.0, with the caveat that the underlying source records remain subject to the license terms of their providers (Europe PMC, NIH, OpenAlex/OurResearch, and the registries listed in the source-manifest summary).