Reproducibility package (I-20): Clusters, not records: effective sample size and the failure of standard interval procedures on irregular instrument channels

Analysis code and computed evidence store for the manuscript "Clusters, not records: effective sample size and the failure of standard interval procedures on irregular instrument channels", submitted to Chemometrics and Intelligent Laboratory Systems.

This deposit contains code and derived results only. The underlying telemetry is published separately under its own DOIs, which fetch_data.py downloads and checksum-verifies before analysis, so the published figures are reproducible from the public record.

Study design: observational audit of already-deposited telemetry (seven recording sessions). No experimental control, no intervention, no independent replication, and no statistical power for population-level inference. Headlines formulated after measurement rather than pre-declared in config.json are labelled exploratory in the manuscript.

Multiplicity: this deposit is one of 15 analyses drawn from the same corpus of seven sessions, one vehicle and one rider. No correction for multiple comparisons is applied across that family, so the family-wise error rate is not controlled.

Repository commit of record: 37f7b2147b26