Companion simulation code and outputs for "Certified de-mixing for direction-locked paired measurements: a metrological statistic with an uncertainty budget, with a ferroelectric domain-wall instantiation"

Reproduction package for the manuscript of the same name (submitted to Metrologia, 2026). Every number, table and figure in the paper is regenerated by the five self-contained Python scripts (numpy/scipy/matplotlib, seeds 20261016–20261018): - sim_single.py — sealing certificates, NAS identity check (10^6 draws), dictionary-misspecification leaks, single-pair suite with oracle-GLS and plug-in comparison- sim_ensemble.py — full-GUM-budget coverage (68%/95%), false lock-claim rates at s=0, naive-slope dispersion statistics- sim_ens3.py — the registered ensemble run quoted in Sec. 6 and Fig. 4b (best-subset load selection, BIC penalty, zero-mean and systematic load regimes, K = 2/4/8)- sim_ens4.py — the unstable estimator variant documented in Appendix B (coordinate-descent limit cycles, variance inflation up to 10x)- k2_figs.py — all four figures from the JSON outputs The four JSON files (single_pair, single_full, ensemble2, ensemble3) are the recorded outputs of the registered runs, included so the paper's numbers can be verified without re-execution. See README for the file-to-section mapping and runtimes.