Frequently asked questions about SGR
Real skeptical objections and questions about the author, answered self-sufficiently based on the article text.
Does SGR prove that the Moon affects human behavior?
SGR documents robust statistical associations — reproduced across five independent city datasets — between astronomical triggers (static phases and impulsive d²F/dt² waves) and population-level crime, EMS, and 911 indicators. That said, the association itself is not proof of direct causation in the strict sense: confirming a causal mechanism requires additional molecular-level verification and prospective protocols with individual monitoring (see §7.7, ‘Limitations of causal interpretation’, in the article).
Is this just the full-moon effect?
No. Most prior lunar-influence studies compare ‘lunar phase day vs. other days.’ SGR explicitly separates two independent layers: a static phase layer (the traditional four phases, Branch A) and an impulsive wave layer — jumps in the second derivative of the tidal potential d²F/dt², detected by the ddF detector (Branch B, with static phases partialled out). The two layers manifest in different classes of outcomes.
Are seasonality, weekdays, and holidays accounted for?
Yes. Before effect estimation, every time series is processed by OLS regression with a degree-4 polynomial trend, an annual Fourier basis (3 harmonics), day-of-week dummies, and U.S. federal holiday dummies with ±1-day lags (§3.3 of the article). Effects are estimated on residuals with the full calendar structure already removed.
Are multiple comparisons controlled for?
Yes, at two levels. Within strata (epoch × trigger family × lag), Benjamini–Hochberg (BH-FDR) correction is applied. Across cities, a two-stage random-effects meta-analysis with HKSJ correction (Hartung–Knapp–Sidik–Jonkman) is used — the conservative standard for a small number of cohorts (k = 5 cities, df = 4), which reduces the risk of anti-conservative underestimation of the standard error (§3.8–§3.9).
Can the analysis be reproduced independently?
Yes. The full source code (base.py, gen.py, correlator.py, verdict.py, single_correlator.py, single_verdict.py), a requirements.txt with pinned library versions, and instructions for downloading the raw city datasets are published on GitHub. An archival snapshot with a DOI is deposited on Zenodo. Code is under the MIT license; processed data and figures under CC-BY 4.0.
Who is the author, and why should the data be trusted?
The author is Artem Tysiatskii, an independent researcher and software developer based in Kaliningrad, Russia (ORCID 0009-0006-1974-7894). All seven source databases are public official open-data portals of U.S. cities (NYC, LA, Chicago, Philadelphia, SF). The author declares no competing interests; the work is self-funded independent research with no external financial support (§9.0 of the article).
What are the main limitations of the study?
Among others: (1) data are aggregated at the calendar-day level, precluding evaluation of intra-day shifts; (2) geomagnetic background (K-index) is not included as a covariate; (3) the observed effects are population-level statistics (~3–6% for property crimes, ~10–20% for individual medical indicators), not individual probabilities. The full list of 7 limitations is given in §7 of the article.