Overview
Chebyshev's theorem guarantees a minimum share of any data within k standard deviations of the mean, whatever the shape of the distribution.
Enter the mean, standard deviation and k. Rule code 2 uses Chebyshev's bound. Code 1 shows the normal-curve figure for comparison.
Worked examples
| Inputs | Result |
|---|---|
| Mean (μ)100Std Dev (σ)15k Standard Deviations2RuleChebyshev’s theorem | Interval [μ - kσ, μ + kσ]: [70, 130]Empirical-rule coverage: 95.45%Chebyshev minimum coverage: 75%Selected benchmark: 75% |
Show 2 more examples
| Mean (μ)50Std Dev (σ)5k Standard Deviations3RuleChebyshev’s theorem | Interval [μ - kσ, μ + kσ]: [35, 65]Empirical-rule coverage: 99.73%Chebyshev minimum coverage: 88.89%Selected benchmark: 88.89% |
| Mean (μ)0Std Dev (σ)1k Standard Deviations1.5RuleChebyshev’s theorem | Interval [μ - kσ, μ + kσ]: [-1.5, 1.5]Empirical-rule coverage: 86.64%Chebyshev minimum coverage: 55.56%Selected benchmark: 55.56% |
How it works
At least 1 − 1 ÷ k² of the data lies within k standard deviations, for any k above 1. It's a worst-case bound, so real data usually have more inside the interval. For normal data, the empirical rule gives much higher figures.
The formula
Practical tips
- Within 2 SDs, at least 75% of any data falls between 70 and 130 when μ = 100 and σ = 15.
- Within 3 SDs, at least 88.9%.
- For normal data, the actual share within 2 SDs is about 95.45%.
Frequently asked questions
Why is the bound so low?
It has to hold for every possible distribution, including very unusual ones.
What about k = 1?
The bound gives 0%, so it says nothing useful there.
When should I use it?
When you don't know the shape of the data.