This is the same baseline-rate cluster illusion as the 2D version, plotted spatially instead of as a bar chart. Each grid cell is an independent region sharing one population size and one baseline anomaly rate. Its case count is a draw from a Poisson distribution, and each individual case is then scattered as one point inside its cell footprint — so what you are orbiting is literally the raw event data, not a smoothed heatmap.
expected (λ) = births × rate / 1000
P(observed = k) = e^(−λ) · λ^k / k!
p-value = P(X ≥ observed) = 1 − Σ (k=0 → observed−1) e^(−λ)λ^k/k!
Because points are placed independently at random, ordinary chance alone will sometimes pack several of them close together — from directly above it looks exactly like a "hotspot." The outlined cell(s) are always whichever cluster of adjoining cells currently holds the single highest count, the same worst-case-picking a concerned community would do. The raw p-value for that cluster is often small; the Bonferroni-corrected value (raw p × number of regions scanned) shows what happens once you account for having scanned the whole map before choosing the "surprising" spot.
- Births per region — population under study per cell; larger populations produce steadier, less noisy counts.
- Baseline rate — the shared population-wide anomaly rate (per 1,000 births) every cell draws from.
- Claimed-cluster size — how many adjoining cells get pooled into the outlined "cluster" before testing observed against expected.
- Resample — redraws every cell's events from scratch; watch the visually "worst" spot on the map jump to a completely different location each time, even though nothing about the underlying rate changed.
Real combined major-birth-anomaly rates run on the order of a few cases per 1,000 births. Random spatial clustering of independent low-count events like this is the leading statistical explanation offered for most publicised birth-defect and cancer "clusters" that do not hold up once tested against the population baseline.