Data and materials for: A source-verifiability workflow for reporting-guideline item sets

This dataset contains the post-audit retained-row manual-audit package for a methods/meta-research study of source-verifiability in reporting-guideline item-set sources. This version includes final editorial cleanup, an abstract shortened to remain safely below the journal abstract limit, and a post-finalization independent masked source-verification audit of all 101 update-inclusive retained source-map rows and all 259 non-retained component rows by two independent masked source auditors. For retained rows, each masked auditor confirmed 92/101 rows; conservative dual masked confirmation retained 90/101 retained rows, corresponding to 90/360 component rows overall and 2/60 complete objects. The full non-retained-row masked audit found one dual-confirmed false-negative candidate, yielding 91/360 full-row masked-confirmed component rows and 2/60 complete objects. A locked-packet-only plus full-row masked-confirmation sensitivity yielded 87/360 retained rows and 1/60 complete objects. The package includes the locked route-stratified source frame, independent Stage A source-retrieval returns, locked canonical source packets, independent Stage B strict locator coding returns, source-reopening adjudication workbook, dual-auditor verbatim-excerpt audit workbooks, completed retained-row manual audit returns, completed independent masked-audit returns, merged masked-audit outputs, primary row-level component dataset, endpoint summaries, display tables, audited complete-pass source maps, data dictionary, complete primary-field dictionary, GRRAS-oriented checklist, manuscript files, standalone Figure 1 files, and reproducibility-check scripts. The source-reopening adjudication endpoint was 57/60 complete objects and 350/360 component rows. It was reduced to 17/60 complete objects and 156/360 rows after conservative excerpt screening and dual-auditor audit of all 194 flagged rows, and then to 2/60 complete objects and 101/360 rows after full retained-row manual audit of the 156 screen-negative retained rows. The locked-packet-only sensitivity was 1/60 complete objects and 97/360 rows. Uploaded package file: JCE_SOURCE_VERIFIABILITY_ORIGINAL_RESEARCH_SUBMISSION_PACKAGE_2026-05-28_FINAL_EDITORIAL_CLEAN.zip. Local MD5 at upload: 209f3ee2150236acdc68f3a0be4b1655.