You are on page 1of 2

Chebyshev’s Theorem

This theorem is named after Russian mathematician Pafnuty Chebyshev,


although it was first formulated by his friend and colleague Irenee-Julkes Bienayme.
The theorem was first stated without proof by Bienayme in 1853 and the later proved
by Chebyshev in 1867.

Definition

Chebyshev’s inequality is usually stated for random variables, but can be generalized
to a statement about measure spaces.

Let 𝑋 (integrable) be a random variable with finite expected value 𝜇 and finite non-
zero variance 𝜎 2 . Then for any real number 𝑘 > 0,

Pr(|𝑋 − 𝜇| ≥ 𝑘𝜎) ≤ 1/𝑘 2


1
Only the case 𝑘 > 1 is useful. When 𝑘 ≤ 1 the right hand side 𝑘2
≥ 1 and the
inequality is trival as all probabilities are ≤ 1.

Example 1.

Annual incomes of the individuals in a population are of an unknown


distribution shape, with a population mean of $80,000 and a population standard
deviation $6,000. According to Chebyshev’s Theorem, what percentage of the income
fall in the interval from $58,400 to $101,600?

𝜇 = $80,000 and 𝜎 = $6,000

Our interval takes us a distance of $21,600 in each direction from the mean
($101,600 − $80,00 𝑎𝑛𝑑 $58,400 − $80,000). But 𝑘 represents the number of
standard deviation we are traveling in each direction from the mean.

𝑘 = 𝑑𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑚𝑒𝑎𝑛⁄𝜎 𝑘 = 𝑆21,600⁄𝑆6,000 = 3.6

Therefore, the interval from $58,400 to $101,600 is the interval beginning at


𝜇 − 3.6𝜎 and ending at 𝜇 + 3.6𝜎.

We now substitute 𝑘 = 3.6 into Chebyshev’s Theorem


1 1
𝑘 2 = (3.6)2 = 12.96 𝑘2
= 12.96 = 0.7716
At least (1 − .07716) × 100% = 𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 92.248% of the incomes fall in the
interval. Remember that this means 92.248% 𝑜𝑟 𝑚𝑜𝑟𝑒.

Example 2.

Dollar amount for a population of invoices are of an unknown distribution


shape, with a population mean $6,500 and a population standard deviation of $400.
According to Chebyshev’s Theorem, at least 95.0617284% of the invoices have values
that fall between what two-dollar amount?

This time we have to work backwards using the theorem. In example 1, we


were given the endpoint of the interval and had to find the percentage chance of being
in that interval. Now we are given the percentage chance of being in the interval and
are required to find the endpoints of the interval.

𝜇 = $6,500 𝜎$𝑆400

Next we need to solve for 𝑘 (the number of standard deviations we are going
in each direction from the mean).
1
At least (1 − 𝑘 2 ) × 100% = 𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 95.0617284%.

You might also like