SPOCKSPOCK // CASE #004 // METHODOLOGY AUDIT

The Cobbler Test

A statistical companion to The Silenced Scientists. When the denominator is a nation, every signal looks like noise — and every anomaly vanishes into the crowd. A cohort-adjusted audit of how the cluster has actually been reported.

Filed April 2026

Classification Methodological

Companion to Case File 001

Status Published

Editorial Lede // SpockSpock

Nearly every outlet covering the Silenced Scientists cluster has anchored its skepticism to the same comparison: ten unexplained events set against the backdrop of roughly three million annual American deaths. That comparison is not rigor. It is a category error.

The test that matters is simpler — and older than statistics itself. Before you ask whether something is strange, you have to ask: strange compared to what? Compare the cluster to the right cohort, with the right age and sex adjustments, and the finding changes shape entirely.

§ 1

The fallacy being tested

Nearly every news segment covering the Silenced Scientists cluster — from cable networks to wire copy — has leaned on the same rhetorical crutch: people die every day, or some variation framing the cluster against the 2.9 million Americans who die annually from all causes. That comparison is not skepticism. It is a category error.

The correct comparison is a cohort-adjusted baseline: what is the all-cause mortality rate for people with the same occupation, the same age distribution, the same gender composition, the same socioeconomic status? Only then can you see whether any given cluster is elevated, suppressed, or unremarkable.

Fortunately, this number exists. It was published in JAMA Surgery in early 2025. No one covering the scientist cluster appears to have cited it. This case file runs the comparison the press has not.

§ 2

At a glance

Summary of findings

The whole-population comparison is unsound. Comparing the Silenced Scientists cluster to the general US population — as most press coverage does — is an analytical dodge that answers no useful question.
Run correctly, the cluster is not a mortality anomaly. The baseline for the lawyer/engineer/scientist cohort predicts roughly 556 deaths in a 75,000-person subset over 22 months. Reported count: 10.
The anomaly is in cause-profile, not count. Zero of ten reported cases fit the five causes (cancer, heart, accident, stroke, suicide) that should account for roughly 85% of baseline deaths.
Unresolved disappearances are the strongest statistical signal. Observed-to-expected ratio ≈ 27×, comfortably outside Poisson noise. This is the finding reporters should be pressing on.
Gender composition is consistent with the underlying workforce and does not drive the pattern.

§ 3

The correct baseline

A Harvard-led analysis of the 2023 U.S. National Vital Statistics System — covering 1,080,298 decedents between ages 25 and 74 — produced age- and sex-adjusted mortality rates for four occupational groupings. Three of them are worth carrying in your head.

Mortality by Occupation

Per 100,000 population, age 25–74, age- and sex-adjusted to the 2000 U.S. standard. Source: Patel et al., JAMA Surgery, 2025.

Two things jump out. First, the lawyers, engineers, and scientists cohort — the correct pooled comparison group for the Silenced Scientists cluster — dies at 404.5 per 100,000, measurably lower than the average working-age American (632.5/100k). High education and high income are the single biggest survival advantages in American life.

Second: the leading causes of death in this cohort are cancer, heart disease, accidental death, stroke, and suicide — in roughly that order. Homicide does not crack the top five. Hold onto that. It becomes the key to the cluster.

§ 4

Running the calculation

Apply the 404.5/100k baseline to three candidate denominators — each a plausible definition of "US scientists with defense, nuclear, or aerospace exposure" — over the 22-month window from July 2024 through April 2026.

Denominator cohort	Population	Expected (22 mo)	Cluster count	O ÷ E
US physicists & astronomers (narrow)	18,732	≈ 139	10	0.07×
+ Aerospace engineers, materials researchers	~42,000	≈ 311	10	0.03×
Est. US classified-access defense/nuclear/aerospace scientists	~75,000	≈ 556	10	0.02×

Read that last row again. If the cluster cohort were dying at the same rate as comparable lawyers, engineers, and scientists nationally, we would expect to count roughly five hundred and fifty-six deaths across the 22-month window. Reported: ten.

Which forces an uncomfortable question for anyone convinced the cluster proves something. By raw count, the cluster is statistically suppressed, not elevated. The whole-population comparison so common in press coverage is hollow. But running the comparison properly actually weakens the simple "they're being killed off" narrative.

§ 5

So why does the cluster still feel wrong?

Because the anomaly isn't in the count. It's in the cause profile. And for that, you have to decompose the 404.5 — ask what those deaths normally look like, and ask whether the cluster's ten look the same.

§ 6

Where the anomaly actually lives

Among lawyers, engineers, and scientists nationally, the distribution of causes of death is extremely stable year-to-year. Compare that stable baseline against the cluster's publicly reported causes and the divergence is immediate.

Expected

Baseline cohort · JAMA 2025

Cancer (neoplasms)≈ 40%
Heart disease≈ 25%
Accidental deaths≈ 8%
Stroke≈ 5%
Suicide≈ 4%
Homicide< 1%
Missing-person (never found)≈ 0.1%

Observed

Cluster · July 2024 – April 2026

Cancer0
Heart disease0 reported
Accidental deaths0 reported
Stroke0
Suicide0 confirmed
Homicide (stranger-perpetrated)2
Disappearances, unresolved5
Cause undisclosed / pending3

The mismatch is near-total. Not a single cluster case matches the five causes that should collectively account for ~85% of deaths in this cohort. Instead, every reported case falls into categories that baseline mortality predicts at ≤1%.

That is the outlier signal. The deaths are not unusual in number. They are unusual in kind.

§ 7

The three rarities

For each unusual cause-category surfacing in the cluster, divide observed count by Poisson-expected count using the 75,000-person denominator and the 22-month window. The result ranks the anomalies.

Rarity · 01

Unresolved disappearances of employed professionals

27×Observed ÷ Expected

Base-rate for working-age professionals who go missing and remain missing long-term is roughly 0.13 per 100,000 per year. Expected in cohort: ~0.18. Observed: 5. This is the single strongest statistical signal in the cluster — and the one least discussed by mainstream coverage.

Rarity · 02

Stranger-perpetrated homicides of research scientists

~3×Observed ÷ Expected

Homicide rates in the professional/scientific cohort run near 0.5 per 100,000 per year, and a majority are known to the victim. Stranger-homicides of scientists (Loureiro, Grillmair, both shot within 60 days) exceed Poisson expectation but remain within the 95% tail of random variation. Suggestive, not conclusive.

Rarity · 03

Geographic concentration at three facilities

N/AInstitutional clustering

Of the ten reported cases, at least seven are tied to three institutions — NASA Jet Propulsion Laboratory (4), Los Alamos National Laboratory (2), and the Kansas City National Security Campus / Albuquerque-adjacent facilities (1+). Cluster-per-facility Poisson math is unreliable at N=2, but the concentration warrants explicit tracking in Case #001.

§ 8

The gender check

Sex is the single largest predictor of occupational all-cause mortality in most BLS data — which is why the JAMA 2025 baseline used earlier explicitly age- and sex-adjusted its figures to control for it. Still, it's worth checking directly whether the cluster's gender composition is itself driving the pattern, or simply tracking the underlying workforce.

Deceased (5): Maiwald, Hicks, Loureiro, Grillmair, Thomas — 5 men, 0 women
Missing (5): Chavez, Reza, Casias, Garcia, McCasland — 3 men, 2 women
Combined: 8 men, 2 women · 4 : 1

US physicist/astronomer workforce: ~81% male, 19% female. The cluster's 4:1 ratio tracks the underlying workforce almost exactly. Gender composition of the cluster is not anomalous — which is why sex adjustment doesn't change the picture, and why the disappearance rate remains the headline anomaly.

§ 9

Verdict

The press has been reporting this story against the wrong denominator. Run correctly — the cohort-adjusted comparison, the age- and sex-adjusted baseline — the data says something stranger than "an unusual number of scientists are dying."

It says: an unusual number of scientists are disappearing without resolution, at a rate that looks roughly twenty-seven times what baseline should produce, while the death count itself remains below baseline. That is the shape of the actual anomaly.

Whether the explanation is foreign intelligence, domestic targeting, a selection artifact from how these cases entered public attention, or something outside the conventional set — Case #001 remains open. But Case #004 closes with a narrower question handed back to reporters and federal investigators: where are the missing, and why is that particular number so large?

§ 10

Methodology & sources

How the numbers were built. Estimates flagged as approximate should be treated as order-of-magnitude figures; the headline findings are robust across any defensible choice of denominator within the stated range.

Baseline mortality

Age- and sex-adjusted all-cause mortality rates (ages 25–74, per 100,000 population, direct standardization to the 2000 US standard) are taken from the National Vital Statistics System via Patel et al.'s analysis.

Patel VR, et al. Mortality Among US Physicians and Other Health Care Workers. JAMA / JAMA Surgery, 2025. Decedent data from the 2020–2023 National Vital Statistics System; population denominators from the American Community Survey. pubmed.ncbi.nlm.nih.gov/39992637

Workplace fatality comparison

US Bureau of Labor Statistics, National Census of Fatal Occupational Injuries in 2024, USDL-26-0230, 19 February 2026. All-worker rate: 3.3 per 100,000 FTE; transportation/material-moving: 12.5; construction/extraction: ~13; refuse collection: 37.4. bls.gov/news.release/cfoi.nr0.htm

Cohort population estimates

US physicists & astronomers workforce (18,732 in 2023, 81.4% male), DataUSA/ACS occupational profile SOC 192010.
Aerospace engineers + materials researchers + related specialties pooled from BLS Occupational Employment and Wage Statistics, May 2024.
"Classified-access defense/nuclear/aerospace scientists" estimate (~75,000) is a rough ceiling built from: DOE national-lab scientific workforce (~28,000 across Los Alamos, Sandia, LLNL, Oak Ridge, et al.), NASA-JPL technical staff (~5,000), Air Force Research Laboratory (~10,500), and major defense contractors' R&D populations (~30,000). Primary citation pending refinement — this estimate should be treated as an order-of-magnitude placeholder, not a precise figure.

Cluster cases

Case list cross-referenced from the House Oversight Committee (Rep. Eric Burlison) public letters, NewsNation reporting, CBS News, Newsweek, CNN, and the Case #001 primary-source index. As of April 2026: 10 cases actively flagged by the White House review; one additional (Hicks, 2023) adjacent but outside the primary cluster window.
Homicide and missing-persons baselines drawn from FBI Uniform Crime Reporting (2023 final) and NamUs historical resolution statistics. Base rates expressed as professional-cohort equivalents rather than general-population rates.

Uncertainty & limitations

The core finding — that cohort-adjusted mortality is below baseline while disappearance rate is 25–30× baseline — is robust under any defensible choice of denominator in the range 18,000 to 150,000. Changing the cohort estimate within that range shifts the ratios but not their sign or order of magnitude. The 75,000 figure used for headline calculations is deliberately conservative; the narrower the cohort, the more suppressed the death rate looks and the more elevated the disappearance rate becomes.