Beware Regional Scatterplots

[Epistemic status: not original] A statistics-literacy note on why regional scatterplots (states/countries) mislead: clustering/spatial autocorrelation. The vivid rainfall/gender example (r=0.84 but really 'three data points' — female-skewed rainy South, male-skewed dry West, and the Rest — so re-running with three cluster-means gives p=0.3), and the income/happiness cluster breakdown. Warns basic scatterplots don't control for this so most are suspect, but also that cluster-choice gives too much wiggle room to prove anything.

Quality 64 · Strong. 64 — low Strong. A clear, useful statistical-literacy contribution; the 'not 48 data points, three data points' reframing of spatial autocorrelation is memorable and practically useful. Held at low-Strong because it's short and Scott flags it as 'not original.'

Claude’s paradigm shift 44 · Moderate. 44 — Moderate. Spatial autocorrelation is known; the clear popularization + memorable example is mildly generative.

Real-world impact 2 · Minor. 2 — minor/within-blog. Statistical-literacy point; discourse-level, no material reach.