Accessible but unfindable: data availability statements fail to name the datasets they exist to disclose

Complete replication package for an audit of data availability statements in cancer bioinformatics, covering all 5,744 MEDLINE-indexed articles published in 2024 that named a public data repository in the title or abstract. Contents of the archive: - out/frame_ids.txt — the sampling frame: 5,744 Europe PMC identifiers, one per line- SEEDS.md — the three fixed random seeds and the script line that sets each one- out/sample_extracted.csv — the 300 sampled records with abstract-level accession coding- out/sample_with_journal.csv — the same 300 records joined to journal, DOI and PMC metadata- out/fulltext_verification.csv — the 162-paper stratified subsample with retrieval outcome and full-text accessions- out/FINAL_counts_dom.csv — the authoritative per-paper coding behind Section 3 of the manuscript- out/manual_check_sheet.csv and out/manual_classification_v2.csv — the 30-paper semantic-validation subset and its manual verdicts- out/parser_crosscheck.csv — DOM-based versus regular-expression extraction of data availability sections- CODEBOOK.md — definition of every variable in the coded dataset- MANIFEST.sha256 — SHA-256 checksums for all 33 files in the package- 10 extraction and analysis scripts, 3 JavaScript audit scripts, and 5 deprecated scripts retained because two of them are discussed in the manuscript- README.md — the order of execution, in four stages- LICENSE (CC BY 4.0) and CITATION.cff The archive contains 34 files. Extract it before use: the manifest and the README refer to paths relative to the archive root, and the scripts resolve paths from their own location. The 112 analysed full texts are not redistributed because they are publisher-copyrighted. All are open access at PubMed Central; the PMCID of each is recorded in out/fulltext_verification.csv, and cache_xml.ps1 re-retrieves all 112 texts in under two minutes. Two independent audit scripts re-derive every reported statistic and confidence interval in JavaScript from the same CSV files, without re-running the PowerShell pipeline, and both pass on the extracted archive.