Before Sputnik · source record · preprint

Independent Replication of Nuclear Test-Transient Correlations and Earth Shadow Deficit in POSS-I Photographic Plates

Doherty, B. · 2026

AI-verified citationSRC-DOHERTY-2026

identifier / locator: arXiv: 2604.00056 (March 2026)
url: https://arxiv.org/abs/2604.00056
verification note: Full text read via arXiv 2604.00056 (2026-08-22 batch).
reliability notes: Independent statistical replication with independently written code — but performed on the original author-supplied 107,875-row catalog, so it tests the analysis, not the source extraction.
Independent researcher; later a co-author of the Bruehl et al. 2026 machine-learning paper, so the two are not fully independent of each other.
Sequence clarified 2026-08-24 via correspondence from the proponent group: this author was initially unconvinced by the transient claims, produced this replication from that starting position, and subsequently joined the group while continuing to work semi-independently. That order matters in both directions - it strengthens the replication itself, which was performed from outside and while skeptical, and it means later output under this name is group-affiliated rather than independent. Volunteered by the proponent side, against interest.

Evidence drawn from this source

Supports· moderateTRN-E010

what the source states: Using the original 107,875-object dataset supplied by the authors and independently written analysis code, the nuclear-window contingency result reproduces (relative risk 1.45, p = .011), survives negative binomial regression with precipitation, lunar-illumination, and cloud-cover controls (IRR 1.80 all transients; 3.98 sunlit-only), and passes a 10,000-iteration permutation test (p = .006); the Earth-shadow classification finds 0.46% of sources in the umbral cone (0.45% in a center-of-plate subset) against a geometric expectation of ~1.4%.
what we infer: The published correlations are not coding errors unique to the VASCO authors — a genuine and useful replication, but of the analysis only. It starts from the authors' own candidate rows, so it cannot speak to whether those rows describe real sky events.
limitations: Independent statistic, not independent extraction: the catalog under analysis is author-supplied (contrast TRN-E009).
Preprint; single author; later co-authored the Bruehl et al. 2026 ML paper, so these two records are not fully independent of each other.