data/explanatory_report.md

Nuclear/UAP Explanatory Factors Private Preview

The better explanation surface is population/reporting geography plus ordinary aviation proximity, while the nuclear-specific proximity claim remains unsupported after controls.

What changed

• Missing geocodes are counted and routed; they are not silently filled.
• Population and major-airport proximity are measured as confounders.
• The previous nuclear matched-control test is carried forward as the nuclear-specific check.

Strongest explanations we can support

• Population explanation strength score: 78/100.
• Airport confounder strength score: 90/100.
• Nuclear non-support strength score: 95/100.
• Place-level log(population) vs log(report count) correlation: 0.641288.
• Events matched to Census population rows: 105086 of 105250.
• Median event distance to a major scheduled airport: 13.277 miles.
• Population-weighted median place distance to a major scheduled airport: 10.764 miles.

Missing geocodes

• Total U.S. rows scanned: 123013.
• Rows without Census-place geocode: 17783 (0.144562).
• These rows need county, landmark, route, relative-place, or fuzzy public-gazetteer recovery before entering spatial tests.

What We Can Claim

• The missing NUFORC-derived rows were not geocoded by assumption; they were excluded from spatial tests and logged for recovery.
• The public geocoded report rows can be compared against Census place population and commercial-airport proximity as explanatory covariates.
• At the place level, report counts can be assessed against population concentration without claiming reports are true or false.
• Airport proximity can be treated as a public-source confounder, not as proof that airplanes caused specific reports.
• The nuclear power-plant proximity claim remains unsupported in this evidence surface after matched non-nuclear power-plant controls.

What We Cannot Claim

• This dataset cannot prove what witnesses saw.
• This dataset cannot claim aircraft explain every report.
• This dataset cannot treat city/place centroids as exact sighting coordinates.
• This dataset cannot rule out all nuclear-weapons, military, or classified-site hypotheses.
• This dataset cannot assume the skipped missing-geocode rows would preserve or reverse the result without a separate recovery run.