Supplementary data and code: Estimating measles vaccine effectiveness from Japanese public surveillance data with quantitative bias analysis

Supplementary data and code for the study Estimating measles vaccine effectiveness from Japanese public surveillance data with quantitative bias analysis. This archive is the same file submitted as supplementary material with the manuscript. Integrity Supplementary_data_and_code.zip — 4,181,851 bytes SHA-256: 8332d2d78756020fbdc9cbdd69b63ec923b41c87da9297a96ebfb0c87a9e8d5a MD5: 31a3908a68718af4555bfd7e321afdba Contents (82 files) code/ — 36 Python scripts: the screening-method vaccine-effectiveness engine, profile-likelihood intervals, quantitative bias analysis (missingness q, selection K, residual/ascertainment B), population weighting, the registered 13-condition coverage simulation, interval end-point audit, independent re-computation with statsmodels, reconciliation of the two independent extractions, and the test suites. out/ — published results: year- and stratum-level estimates, scenario fits, leave-one-year-out, dose-specific and dose-allocation tables, coverage simulation output, reproducibility comparisons, QC-trigger and small-sample determinations. restricted/ — the fixed extraction signed by both extractors (case counts by year, age and vaccination category), the Q2/Q3 grading, and the result tables that carry the underlying counts. These are aggregate counts reconstructed from published national surveillance charts; they contain no individual-level data. sources/ — the source manifest with URLs and SHA-256 for every public input, plus the population estimate workbooks used for weighting. README.md — environment, the two reproduction paths (re-analysis from the fixed extraction; re-extraction from the public sources), and a key to the Japanese column names. Reproduction Tested with Python 3.14.4, numpy 2.5.2, pandas 3.0.5, scipy 1.18.1, openpyxl 3.1.5 and statsmodels 0.15.0. Path A re-runs the full analysis from the fixed extraction without any download; Path B re-downloads the public sources and repeats the chart extraction. The second, independent extraction and the reconciliation involved human review and cannot be re-run automatically; they are documented in the registration. Determinations fixed before vaccine effectiveness was computed qc_trigger.py and small_sample_layers.py refuse to recompute the determinations that were fixed in advance; --recompute is required to reproduce them. Preregistration Protocol and statistical analysis plan version 2.4, with the change log and deviation record: https://osf.io/aefwp