Generator

spread

A generator in the probability library, called 7 times across 2 essays. Below: what it draws at its defaults and at each mode an essay asks for, what it checks while drawing, and everywhere it is used.

spread is one function. Everything below came out of it during this build, at parameters taken from the essays rather than invented for this page — so a figure here is the same figure a reader meets in an essay, and if the generator changes, this page changes with it.

At its defaults

five values, unevenly weighted, and the mass outside 3 standard deviationsA distribution drawn as bars, with the windows one and a half, two and three standard deviations wide marked. The probability outside each window is summed and compared with the bound that knows only the variance.-6-4-202460.000.100.200.300.400.50valueprobability1.5σthe meanmean 0.300, standard deviation 1.847; beyond 1.5 of them lies 0.2000 against a bound of 0.4444, beyond 2 ofthem lies 0.1000 against a bound of 0.2500, beyond 3 of them lies 0.0000 against a bound of 0.1111the bound knows only the variance — not the shape, not the number of values, not whether the distribution issymmetric — which is why it is so far from tight here and cannot be improved in general

show: "tight"

The distribution that meets the bound at 2 standard deviationsA three-valued distribution whose mass beyond k standard deviations is exactly the bound. It shows the inequality cannot be improved without assuming more than a variance.−2σ0.1250the mean0.7500+2σ0.1250three values, at −2, 0 and 2 standard deviations, with weights 0.1250, 0.7500, 0.1250the mass at 2 standard deviations or beyond is 0.2500, which is exactly 1/2² — so no bound of thiskind can be lower, whatever else is assumed about the shape

show: "sweep"

Four distributions, and the one bound over all of themThe probability of landing at least k standard deviations from the mean, for four distributions, against the bound that knows only their variance. Every staircase stays under the curve.0.511.522.533.540.000.200.400.600.801.00window, in standard deviationsprobability outside itthe bound, 1/k²a fair coin, ±1a fair die0 nine times in ten, 10 otherwisefive values, unevenly weightedthe bound is the falling curve; every distribution's own tail mass is a staircase under it, because a finitedistribution's mass jumps as the window passes a valuethe lopsided one reaches 89% of the bound, and the three-point distribution of the previous figure reaches all of it— so the bound is not merely true but as low as a bound can be

show: "average"

7 running averages, and the band they stay insideThe running average of repeated draws, for several independent runs, with the band two standard errors wide. The runs wander early and settle onto the mean as the band closes.020040060080010002.04.06.0draws so faraverage so farthe mean, 3.507 runs of up to 1,000 draws from a fair die; the band is ±2σ/√n, drawn from the distribution rather than fitted to therunsthe average settles because the band shrinks, and it shrinks like one over the square root of the count — so tentimes the work buys about three times the precision

show: "scaling"

One set of sums, two scalings, two different limitsThe exact distribution of a sum of n independent copies, scaled two ways. Divided by n it collapses onto the mean; divided by the square root of n it holds a fixed width and settles into a shape.-1-0.50.510.01.02.03.0the average, minus the meann = 4n = 16n = 64-3-2-11230.00.10.20.30.4the sum, minus n means, over √nn = 4n = 16n = 64the same exact distributions, drawn twice as densities: divided by n the spread falls 0.500 → 0.250 →0.125, divided by √n it is 1.000 every timeso the left curves climb — 0.75 → 1.57 → 3.18 at the peak — and the right ones settle on 0.399, which is theheight of the bell curve

What it checks while it draws

Collected by running the family and recording what it asserted, not written here. The count is how many separate times the claim was put to the test while these drawings were made.

Where it is called

Changing this generator changes every figure on this list. That is what makes the list worth publishing rather than keeping in a check script.

The whole library · What the figures prove