Regression Diagnostics: Leverage & Cook's Distance

A regression line's coefficients are averages, but they aren't democratic — a single point far from the center of the data can steer the fit far more than the rest of the cloud combined. This simulator renders an ordinary-least-squares scatter in 3D, where each point's height above its base marker encodes its Cook's distance, recomputed on every frame as you drag points around or add a deliberately extreme outlier. Live readouts break the diagnostic into its two ingredients — leverage (how unusual a point's x-position is) and standardized residual (how far it falls from the line) — so you can see exactly why a point is or isn't influential, not just that it is.