Inspection Paradox: Why You Always Wait Longer Than Average (Length-Biased Sampling)
Buses arrive with Gamma-distributed gaps and passengers show up at random times. Watch passengers land preferentially in long gaps, so the gap they experience and the wait they endure exceed the timetable average. Tune the regularity of service and compare the live histograms with E[X squared] over E[X] and the residual-life density, with Monte Carlo and quadrature checks built in.
Four small datasets, laid out here as four glowing boards in a 3D gallery, share nearly identical mean, variance, correlation and least-squares regression line — and look completely different the moment you actually look at them. This is the classic 1973 demonstration by statistician Francis Anscombe of why summary statistics alone can never replace a plot: a linear trend, a smooth curve, a near-perfect line broken by one outlier, and a correlation created entirely by a single extreme point all produce the same numbers on paper. Orbit the scene to compare all four boards at once, focus any dataset to read its live statistics, and drag the jitter slider to perturb the points in real time and watch the correlation, regression line and R² recompute on the fly.
Buses arrive with Gamma-distributed gaps and passengers show up at random times. Watch passengers land preferentially in long gaps, so the gap they experience and the wait they endure exceed the timetable average. Tune the regularity of service and compare the live histograms with E[X squared] over E[X] and the residual-life density, with Monte Carlo and quadrature checks built in.
3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install