About the project
A field guide to being honestly wrong.
unspurious explains the statistical illusions that turn correct numbers into confident mistakes — the ones that decide verdicts, headlines, league tables and diagnoses.
What this is
The premise
Most guides to "lying with statistics" are really about lying — fabricated data, cherry-picked samples, doctored charts. unspurious is about something more unsettling: the times nobody lied at all. Every illusion in this compendium is built from numbers that are individually, checkably true, and yet leads a careful reader to the wrong conclusion. The error lives not in the data but in the silent step from number to interpretation.
The collection is organised by mechanism rather than by field, because the same trap reappears everywhere wearing different clothes. Simpson's paradox looks identical whether it surfaces in a hospital, a courtroom or a baseball box score; once you can recognise the shape, you can spot it anywhere. Each entry opens with an interactive figure where the illusion happens in front of you, walks through a real historical case, and closes with the single question that would have caught it.
How the figures work
Everything here is reproducible
Every chart on the site is drawn live in your browser from either documented historical figures or a clearly-labelled, seeded simulation — never a static image massaged to make a point. The interactive tools run genuine statistics: the base-rate explorer computes real confusion matrices, and the p-hacking sandbox runs an actual Welch t-test with a proper Student-t distribution behind every p-value. If a figure claims a number, you can move the dials and watch that number hold up.
This matters because the whole thesis of the site is that honest arithmetic can still mislead. It would rather undercut that argument to demonstrate it with dishonest arithmetic.
A word of caution
What this is not
unspurious is an educational project about statistical reasoning. The medical, legal and financial examples are illustrations of how to think, not advice for any specific decision — a real diagnosis, verdict or investment depends on details no general explainer can see. Where the site uses simplified or seeded data to make a mechanism visible, it says so plainly.
If you spot an error, an unclear explanation, or a paradox the collection ought to cover, that feedback is genuinely welcome — the contact page is the place for it. The compendium grows as entries are written, and its roadmap is shaped partly by what readers ask for.
Start here
Three good ways in
Browse every entry
All the illusions, organised by mechanism, with a one-question checklist at the bottom.
№ 1 · A CLASSICSimpson's Paradox
Every group says one thing; the total says the opposite. The cleanest place to begin.
A TOOLThe Base-Rate Explorer
See why a 99%-accurate test for a rare condition is usually wrong when it says yes.