Residual Divergence Map¶
Metric or dashboard — instantiates Correspondence Violation Detection and Theory Refinement
Displays, across the whole state space at once, where and how much observed behavior diverges from expected correspondence, read against the error budget.
A Residual Divergence Map is a visualization that displays, across the whole state space at once, where and how much observed behavior diverges from expected correspondence — a field of residuals rendered so the spatial or structural pattern of disagreement is visible at a glance. Its defining quality is that it is a display of the already-computed divergence field over the entire domain, read against the error budget so that coherent structure can be told apart from noise. It does not actively search input grids for a breakpoint, and it does not carry a single case's dossier — it shows the shape of the disagreement everywhere simultaneously, which is what lets a real localized boundary announce itself as a pattern that an aggregate summary statistic would have averaged away.
Example¶
A new global gravity-field model, derived from satellite gravimetry, should correspond to the older reference geoid. A Residual Divergence Map plots the residual (new minus reference) as a color field over the Earth's surface. Most of the globe is pale — within measurement noise — but two things jump out: a coherent deep-red anomaly over an ocean subduction trench, and a faint regular striping running along the satellite's ground tracks. Read against the error budget overlaid on the map, the trench anomaly stands well above noise and has the compact shape of a real geophysical feature, while the stripes have the tell-tale periodicity of an instrument artifact. The map separates a genuine boundary from a processing artifact by their spatial signatures — a distinction that a single global RMS number would have buried entirely.
How it works¶
- Compute residuals everywhere. For each point in the state space, take observed minus expected and store the residual, building a field rather than a summary.
- Render as a field. Display the residuals as a heatmap, geographic map, or scatter over the state space, so magnitude and location are both legible.
- Overlay the error budget. Superimpose the noise floor or tolerance so that within-noise regions and significant excursions are visually distinct — standardized residuals rather than raw ones.
- Read the structure. Coherent shapes signal real boundaries, periodic patterns signal artifacts, and diffuse scatter signals noise; the map feeds triage and per-case cards.
Tuning parameters¶
- State-space axes and projection — which dimensions are shown and how. The right projection makes a boundary obvious; the wrong one hides it in a collapsed dimension.
- Residual metric — absolute, relative, or standardized (divided by the error budget). Standardized residuals separate signal from noise honestly; raw residuals can exaggerate high-magnitude but statistically-insignificant regions.
- Color scale and thresholding — how residual magnitude maps to visual salience, and where the "within noise" cutoff sits against the error budget.
- Aggregation resolution — how finely the space is binned. Fine bins reveal thin structure but get noisy; coarse bins are stable but can smear a localized failure.
When it helps, and when it misleads¶
Its strength is that it defeats the archetype's aggregate-agreement-masking-local-failure trap: a model can pass on average while breaking in one region, and the map makes that region visible as a shape instead of letting a good headline metric hide it. It is the mechanism that turns "overall it agrees" into "except right here, in this pattern."
Its failure mode is over-reading the picture — mistaking a rendering or scaling artifact for a real boundary, or seeing structure in what is actually noise, the visual analogue of finding faces in clouds. Anscombe's quartet[1] is the standing warning in the other direction: identical summary statistics can hide wildly different structures, which is exactly why the map exists, but the same power to reveal pattern makes it easy to perceive pattern that is not there. The guarding discipline is to always render residuals against the error budget rather than raw, and to confirm any apparent structure statistically before treating it as a violation rather than a coincidence.
How it implements the components¶
The map fills the show-the-whole-field part of the machinery:
divergence_signature— the rendered field is the signature: where the divergence sits, how large it is, and what shape it takes across the state space.uncertainty_and_error_budget— the residuals are displayed against the noise floor / tolerance, so the map operationalizes the error budget by separating significant excursions from noise.
It does not implement overlap_or_limit_regime_set — defining and searching the regime axes where correspondence is probed is the job of Parameter Sweep Matrix. The sweep actively searches input grids to find the breakpoint; the map displays the residual field already computed over the state space and reads it against the error budget.
Related¶
- Instantiates: Correspondence Violation Detection and Theory Refinement — the map is the whole-domain view that reveals where correspondence breaks as a pattern.
- Sibling mechanisms: Anomaly Triage Board · Assumption Audit Worksheet · Boundary Condition Memo · Expert Adjudication Panel · Failure Mode Annotation Card · Limiting-Case Test Suite · Overlap-Regime Benchmark Table · Parameter Sweep Matrix · Regression Correspondence Harness
Editorial Notes¶
Form Classification¶
Form family: Monitoring, Sensing & Alerting
Rationale: Residual Divergence Map operates by repeatedly computes observed-minus-expected residuals across state space and renders the changing field. That concrete deployed or enacted form is Monitoring, Sensing & Alerting under the frozen taxonomy.
Nearest alternative: Interface, Display & Cue — Although Interface, Display & Cue can support this mechanism, the frozen evidence makes its operative form the act that repeatedly computes observed-minus-expected residuals across state space and renders the changing field; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Data Science & Analytics
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: State-space visualization of observed-versus-expected error is characteristic of modern data analytics.
Related originating lineages:
- Statistics & Experimental Design — Residual diagnostics and error budgets materially supply the inferential comparison.
Review resolution: Both blind reviewers agree that data_science is the primary historical origin. Explicit reconciliation of alternate origin disagreement adopts reviewer_a's evidence: State-space visualization of observed-versus-expected error is characteristic of modern data analytics. The selected record uses alternates=statistics_experimental_design, origin_mode=cross_disciplinary_synthesis, and domain_reach=multi_domain; the other review proposed alternates=mathematics, statistics_experimental_design, systems_cybernetics, origin_mode=cross_disciplinary_synthesis, and domain_reach=multi_domain. The selected combination better preserves the mechanism-specific formative lineages and calibrated scope; broader present-day use is not treated as proof of additional historical origin.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Anscombe's quartet (Francis Anscombe, 1973) is a set of four datasets with nearly identical summary statistics — mean, variance, correlation, regression line — yet strikingly different when plotted. It is the canonical demonstration that aggregate numbers can hide structure a picture reveals, which is precisely the failure a residual map is built to catch. withdrawn registry ↩