Incoprea Research
Measurement, datasets and the places where numbers stop meaning what they claim.
Work published here is built to be checked: every result links to its data, its script and its
limitations. Negatives are published alongside positives, and claims that failed their own tests
are recorded rather than withdrawn quietly.
Current work
Magnitude-scale mixing and the apparent stability of the b-value
The pooled Gutenberg–Richter b-value is stable to ±2% across three years while the per-scale
values span 0.70 to 2.23 — a factor of 2.3 to 3.2. The pooled number belongs to none of the five
populations in the file. Includes a floating-point threshold error that shifted b by 28%.
Capabilities and services
What I do, the artefacts that demonstrate it, and how engagements are scoped. Small
fixed-scope audits are the usual starting point.
Method
Three rules apply to everything on this site. Every number traces to a file that ships with it.
Every estimate is labelled as an estimate. And when my own test contradicts my own claim, the
correction is published — the collection currently carries fourteen such corrections.
libraries — hundreds of thousands of structured questions. They are labelled as questions, not
findings, and are not presented as results. The findings are the small number of items with data
and tests behind them.