Statistics

Central Limit Theorem

Take a wildly non-Gaussian distribution, draw n samples, average them, and record the mean. Repeat. No matter the source, the distribution of those means tends to a bell curve, and it tightens as n grows.

Means collected0
Std of mean

Faint bars at the top are the source distribution. The big histogram is the distribution of sample means, with the predicted Gaussian overlaid. Set n = 1 to see the raw source, then raise n and watch it turn into a bell.

What to observe

  1. Set n = 1: the big histogram is just the raw source, so Exponential and Bimodal look nothing like a bell.
  2. Raise n to 5, then 30. However skewed the source, the means pile into thesame Gaussian shape. That universality is the theorem.
  3. Watch the Std of mean readout as n grows: it shrinks like1/√n, so the bell gets narrower but four times the samples only halves the width.

Shortcuts: space run/pause · s step · r reset · f fullscreen