UNCERTAINTY STUDIO

Make the uncertainty model inspectable.

METRAXIOM separates the current controlled Decision Engine from richer reference uncertainty work. The controlled engine presently retains standard-uncertainty magnitude, sensitivity coefficient, source/rationale and correlation status; Type A/Type B classification, distribution normalization, degrees of freedom and coverage-factor basis are not yet first-class controlled fields.

Controlled scope stays narrow

Independent standard components may use RSS after sensitivity coefficients are applied. If correlation is declared, the controlled simple-RSS path stops and requires an explicit covariance model.

Reference budget is richer

The composer below exposes Type A/Type B classification, rectangular and triangular Type B normalization, explicit expanded-value divisors, component degrees of freedom and Welch–Satterthwaite effective degrees of freedom.

Coverage is explicit

U = k × u_c. No universal k = 2 is assumed, and this reference tool does not infer a confidence level from a numeric coverage factor.

REFERENCE ONLY · INDEPENDENT COMPONENTS

This composer makes evaluation class, normalization, sensitivity, degrees of freedom and coverage-factor basis inspectable. It does not expand controlled-engine authority. Declared correlation blocks calculation until a covariance-capable model is used.

How to use this reference budget

Type A evaluates uncertainty statistically from observations. Type B uses other information. These are evaluation methods, not synonyms for random and systematic effects.

Enter the evaluated Type A standard uncertainty here. A coverage factor alone does not establish a confidence level. This composer does not automatically save a controlled budget.

NIST TN 1297

Coverage factor k (optional)

Coverage-factor basis

Component 1

Name

Evaluation class

Normalization

Entered magnitude

Divisor / multiplier

Sensitivity coefficient

Degrees of freedom

Source / basis

Correlation

Rationale / note

Normalization ruleStandard uncertainty u

Component 2

Name

Evaluation class

Normalization

Entered magnitude

Divisor / multiplier

Sensitivity coefficient

Degrees of freedom

Source / basis

Correlation

Rationale / note

Normalization ruleStated expanded value (÷ explicit divisor)

Component 3

Name

Evaluation class

Normalization

Entered magnitude

Divisor / multiplier

Sensitivity coefficient

Degrees of freedom

Source / basis

Correlation

Rationale / note

Normalization ruleRectangular half-width a (÷√3)

Open full Workbench Inspect governance boundary Challenge the controlled candidate engine

Automatic counts. No individual tracking.

TRACE counts approved public page requests worldwide without an analytics consent banner. No visitor identifiers, fingerprints, individual browsing histories or GPS locations are collected.

Analytics privacy update · October 29, 2026

TRACE automatically measures aggregate requests to approved public METRAXIOM pages worldwide to improve content, navigation and reliability. It runs at the first-party hosting proxy, before the hosting page cache, and works when JavaScript is disabled. No analytics consent banner or country-based opt-in selection is used.

The basic store contains separate hourly totals for approved pages, country and US-state labels, common referral domains, broad device/browser/operating-system families and unverified request categories. It contains no individual request ledger, full IP addresses, raw user agents, full URLs or queries, visitor identifiers, fingerprints, session histories or GPS coordinates. Country labels come from hosting network metadata and can describe a VPN or network exit. Missing geography is recorded as unavailable. No city, address or coordinates are collected. US-state labels come only from the trusted hosting header; other state/region labels remain unavailable. Software names are broad, spoofable claims, with no versions or cross-request fingerprint.

GPC, DNT and an explicit traffic objection stop collection. Earlier “Keep analytics off” choices saved only in browser storage are carried forward as objections when JavaScript runs on a public application page. They are not sent with the initial HTTP request: that request, JavaScript-disabled requests and static-only visits may contribute until an objection cookie or privacy signal is sent. Exclude this browser from traffic measurement without an account or JavaScript. The preference cookie uses the same value for everyone who excludes traffic (1) or explicitly restores basic measurement (0), lasts up to 180 days and does not identify you. Restoring basic measurement preserves the old local choice; it can apply again after the restore preference expires or is cleared. Restoring aggregate measurement does not override privacy signals or grant individual tracking permission. Earlier aggregate counts cannot be linked back to you or selectively removed.

When you choose a dashboard view, a separate functional reporting-preference cookie remembers only the preset or calendar dates for up to 30 days, scoped to /traffic. It is not used by the collection layer and contains no identifier or consent grant. Use Forget saved reporting view, including without JavaScript, to remove it. A failed refresh may display a clearly dated snapshot held briefly in page memory; it is never relabeled as current or replaced with fabricated zero counts.

Individual browser analytics is disabled in this aggregate-only rollout. The former optional browser collector is not loaded, its intake is retired, and no new tab/session events are accepted. Earlier consent records are not rewritten as new permission. Previously collected optional events follow their existing 30-day retention and integrity-checkpoint policy.

Page requests are separate from verified page rendering or unique people. Navigation requests, crawler claims, link previews, other automation and unknown requests remain separate. Estimated visitors and session histories are unavailable. Small labeled groups are combined in public reports. Counts are aggregate observations and do not establish identity, intent or endorsement.

Recognized health checks, HEAD requests, prefetches, server-component fetches and private/API/unapproved paths are excluded. Browser memory/disk cache, back-forward restores, client-only navigation, privacy objections, missing metadata and storage failures leave coverage gaps. Request retries and spoofed software labels can affect totals. The system does not extrapolate missed traffic or import historical Vercel Analytics counts.

Hourly counter buckets are retained for 30 days and pruned hourly in bounded runs; calendar reports exclude older hourly data even if maintenance is delayed. Separate cumulative marginal totals, with no event times or cross-dimension links, remain for the life of the site’s traffic record. Changing a date range, leaving the page or closing a browser does not erase those totals. Earlier uncollected state/software labels remain unavailable, not reconstructed. Vercel delivers the website and Supabase stores TRACE aggregates in the existing United States database. Provider security logs and backups have separate policies. TRACE adds no paid analytics feed or operator-computer dependency.

This October 29, 2026 supplement updates the analytics description in the September 7, 2026 Privacy Notice. It does not amend frozen agreement snapshots or claim that server-side processing is automatically exempt from privacy law. The assessment concerns this limited data flow and its controls; no universal regulatory approval is claimed. Privacy requests: ANDREWATTS23@gmail.com.