forty hand-picked us regions, twenty indicators, and your priorities —
entered as ranges, not numbers, because you don't actually know
that schools matter exactly 1.3× as much as trails. i monte-carlo your
uncertainty through every ranking. [WIP]
how these places group, before you say anything
plain k-means on standardized indicators — no weights, no opinions.
each card is a family of statistically similar places.
what matters to you
mode
4,000 weight draws · deterministic seed · runs in your browser
results
score = weighted mean of category z-scores, so 0 is "average of these 40."
the bar is the 90% credible interval under your weight uncertainty — when two bars mostly overlap,
the order between them is closer to a coin flip than a fact. click any column header to sort;
click a region for raw numbers and sources.
your top ten, pairwise. each cell: how often the row region beats the column region
across all weight draws. 50% is a coin flip — basically tied. 90%+ is a real gap. click a cell for the score-difference interval.
click any cell to read it in plain words.
reading guide: ≥90% clearly ahead · 65–90% probably ahead ·
50–65% basically tied. these probabilities come from your weight ranges — tighten the ranges and ties can resolve.
the same regions, re-clustered on your weighted category scores —
how your values regroup the map. places in one card are interchangeable under your priorities.
no tiles, no basemap service — census state outlines, baked at build time.
color is an anomaly scale: warm = above the 40-region average, cold = below. click a dot to read it.
dimensionvalues
click a dot.
every region on one dimension. raw = the actual number.
standardized = z-score across these 40 (log-scaled first where flagged).
weighted = z × your mean weight for its category.
dimensionvalues
how the category scores move together across these 40 places
(pearson r on standardized scores; warm = together, cold = opposite).
click any cell to open that pair in the scatter — or pick any two dimensions yourself,
including raw vs standardized of the same one, to see what the transforms do.