RIVET-Bench v1.0: Reference-Interval Benchmark Outputs, Supplemental Data and RIVET-RI Validation Provenance

RIVET-Bench v1.0 is a reproducibility and Supplemental Data archive accompanying a real-world benchmark of indirect reference-interval estimation pipelines using the Standard Biochemical Profile of Laboratory Data from eligible NHANES datasets/cycles. The benchmark evaluated 22 biochemical analytes across 192 defined filtering–estimator pipelines, comprising 4,224 pipeline–analyte opportunities, of which 4,146 produced valid estimates. The archive contains machine-readable comparator reference intervals, pipeline-generated reference intervals, detailed analyte-level reference-limit agreement (“ranging”) and classification-discordance (“misclassification”) results, integrated scores and rankings, cross-source consensus analyses, filtering-stability results, distributional shape and tail characterization, distribution–performance associations, estimability/failure analyses, and finite-sample convergence analyses. The archive additionally provides public validation and provenance documentation for RIVET-RI v1.0, the accompanying hosted application for pooled indirect reference-interval generation. RIVET-RI is available at https://docsrvnyk-rivet-ri.hf.space/. The production deployment passed 15/15 cross-platform numerical validation jobs against the frozen scientific engine with a maximum absolute numerical difference of 0. Raw participant-level NHANES data, row-level filtered datasets, source publications/textbooks, and the protected deployable RIVET-RI source code are not redistributed. Complete archive manifests, SHA-256 checksums, scientific fingerprints, interface documentation, and API usage instructions are included to ensure provenance and reproducibility.