That Wasn't Very Data-Driven Of You

A hundred and twenty-eight points, correlation minus zero point nine. The relationship could not be clearer: the cloud runs downhill, hard. Then someone circles a single point, draws an arrow off it at forty-five degrees in the opposite direction, and labels it n = 1, R² = 1. Three things are wrong with that, and they are wrong in a very particular way. One point cannot determine a line at all — you need two — so the fit is not merely bad, it is undefined. R² = 1 is guaranteed at n = 1, because with one observation the residual is zero by construction, which makes a perfect fit meaningless rather than impressive. And the slope has the wrong sign: the data says minus zero point nine, the finding says plus one. The conclusion contradicts the plot it is drawn on. Nobody here is being fooled. Someone is bluffing, in public, on a chart that disproves them. The dashed line is what the data actually says, printed alongside so you can watch the arrow ignore it. Every number is real: the cloud is a genuine draw from a bivariate normal, the r is the actual sample correlation of the points on the shirt, and the circled point is the one that genuinely sits furthest up and to the right — the best available cherry to pick. For data scientists, analysts, statisticians, and anyone who has sat in a meeting where a decision arrived wearing a chart. Printed in white on dark or black on light.