GUIDES

Short answers.
Stated limits.

Each guide answers one question, shows a worked or recorded example and lists what the method cannot establish.

• GUM vs Monte Carlo uncertainty propagation: when results differ. Both methods propagate the same declared model; the guide shows one recorded case where they agree and one where the first-order budget is inadequate.
• GRIM and GRIMMER checks: arithmetic screening, not fraud detection. A rounded mean can be impossible for integer data at a given sample size; the guide explains the assumptions, two worked examples and why a failed check is not evidence of fraud.
• What a signed calculation receipt proves, and what it does not. A receipt shows origin and integrity within a stated scope; it does not certify the truth of observations or the adequacy of a model.
• Your first AI calculation mission. Connect an assistant, copy a reproducible request in English or French, then inspect a recorded plan and its explicit assumptions.

First published 2 November 2026 · Last updated 2 November 2026