Macao population fitness surveillance summary data and specification-curve analysis, 2005–2020

Abstract

Version 2.0.0 expands the previously published 2,040-record dataset to 2,584 aggregate source records by adding 544 student records. It contains the expanded analysis supporting the manuscript “Divergent population fitness trajectories in Macao, 2005–2020: a specification-curve analysis of repeated cross-sectional surveillance reports.” The primary functional analysis uses 1,544 records and 26 population–test components. Each population contrast has 24 nominal analytical configurations; preschool children have eight distinct formulas. The source statistics come from the official English-language Macao population fitness surveillance reports for 2005, 2010, 2015 and 2020. The archive contains the curated dataset and data dictionary, source-page and table mappings, the candidate-test eligibility audit, Python analysis and plotting code, software requirements, formula-equivalence maps, decision-factor results, sensitivity analyses, figure-source data, generated figures, and the checked data and supplementary-table workbooks with their companion sources. No participant-level or identifiable data, official-report PDFs or journal full texts are distributed. This release includes a refined verification entry point that distinguishes exact output reproduction from differences confined to three recorded environment fields. All scientific-output and protected-input checks remain enforced. Two clean reruns in separate output directories, using Python 3.12.10 and 3.12.14 with NumPy 2.5.2 on Windows, reproduced all 153 statistical CSV outputs byte for byte. The reference-environment run also reproduced all six checked JSON files exactly; the alternate run differed only in the recorded Python version. The statistical calculation modules, numerical results, figures, figure sources and checked workbooks are unchanged by this verification refinement. Current validation evidence and preserved historical verification records are identified separately in the archive. The source dataset for reuse and recomputation is data/macao_fitness_M1.tsv. The README and release manifest distinguish current analysis inputs from frozen regression fixtures and historical workbook provenance fields. UTF-8 reading guidance is provided for files with or without a byte-order mark; existing data-file bytes have been preserved. CC BY 4.0 applies only to the authors’ licensable data curation, harmonisation, original metadata, documentation, analysis outputs (including the generated figure files), and figure-source data. Embedded fonts in figure PDFs retain their own rights; no standalone fonts are bundled. Author-created Python code is licensed under the MIT License. These licences do not relicense the Macao official reports or their original content. The 19 age-band population counts from United Nations Statistics Division, Demographic Yearbook 2021, Table 7, and their direct reproductions remain third-party UN data under the UNdata Conditions of Use. Full scope and attribution details are provided in LICENSE_DATA.md, LICENSE_CODE.txt and THIRD_PARTY_MATERIALS.md.