Explore linked distributions. Compare equipment. Trace every observation to evidence.
Compare model rates on a common scale. Intervals show sampling uncertainty, not reporting bias. Select a model to inspect its reports.
Report-level evidence · select a row to investigate
Every event, model, manufacturer and exposure record is fictional. The seeded generator observes a primary equipment item per dive; denominators describe item-dives, not all equipment on an actual dive.
Click a category, model or geographic cell to add or remove it. Selections within a dimension combine with OR; different dimensions intersect with AND. Pale marks retain context with that view’s own filter removed. Depth and age brushes select an inclusive numerical range; exact inputs are the keyboard alternative. The conditions matrix selects temperature/service pairs, not their independent Cartesian product.
Equipment, manufacturer, model, region, depth, temperature, service and month filters select both events and eligible exposure. Severity and failure mode select only the numerator: rates describe that event subtype per eligible dive. Search, map hexagon and rectangle selections suppress rates because a defensible exposure population is unavailable.
Missing reported depths remain in event counts. Depth and temperature filters use the complete simulated dive log. The map switches between fixed equal-area projected hexagons (count intensity) and screen-space incident clusters (area proportional to count). Hexagon membership does not change when zooming. Cells are a projected analytical grid, not sea boundaries. Dashed outlines indicate approximate coordinates; they are not spatial confidence regions. Coordinates are illustrative; dashed halos indicate approximate reports. Unknown positions remain in other views.
95% Wilson binomial intervals assume independent primary-item dives; repeated assets, reporting bias and confounding are not modelled by these intervals. Differences are descriptive. No causal, predictive or corrective-action effectiveness claim is supported.
D3 choropleth: restrained geographic rendering and a quantitative colour legend. D3 Graph Gallery: histogram brushing, heatmaps and comparable dot plots.
These are design interpretations of public documentation, not an evaluation of restricted dashboards. Coastlines: Natural Earth / world-atlas, public domain. No bathymetry is shown.
Changing seed or exposure regenerates related records and resets selections. Authored signals relate equipment, conditions and service history.