We added Anscombe's quartet to the correlation explorer

Anscombe's quartet is four small datasets with the same mean, the same variance, the same correlation, and the same regression line — and four completely different shapes. One is linear, one is a clean parabola, one is linear with a single outlier dragging the fit, and one is a vertical stack with a single point creating the entire correlation.

It is from 1973 and it remains the most efficient argument ever made for plotting your data.

It is now a dataset here, so you can load it into the correlation explorer and watch four identical coefficients sit above four unrecognisable scatters.

What the tool adds beyond the classic picture is the Spearman column. On set two — the parabola — Pearson is 0.816 and Spearman is 0.691, and both are misleading in different ways. On set four, the vertical stack, Spearman is undefined-ish because ten of eleven x-values are identical. Seeing the two coefficients disagree, or fail, is a second lesson on top of the first.

We also added the Datasaurus dozen while we were at it — thirteen datasets with identical summary statistics to two decimal places, one of which is a dinosaur. It is the same lesson with a better punchline, and it makes the point to people who find Anscombe abstract.

Both are tiny, both are CC0, both are on the datasets page.

If you take one habit from this site, take this one: plot it before you summarise it. The plot takes ten seconds and it is the only thing that distinguishes these four cases.