Overview
The empirical rule, or 68-95-99.7 rule, says how much of a normal distribution lies within 1, 2 and 3 standard deviations of the mean. This gives the interval and the coverage for any number of standard deviations.
Enter the mean, standard deviation and k. Use rule code 1 for the normal-curve figure, or 2 for Chebyshev's guaranteed minimum, which holds for any distribution.
Worked examples
| Inputs | Result |
|---|---|
| Mean (μ)100Std Dev (σ)15k Standard Deviations2RuleEmpirical rule | Interval [μ - kσ, μ + kσ]: [70, 130]Empirical-rule coverage: 95.45%Chebyshev minimum coverage: 75%Selected benchmark: 95.45% |
Show 2 more examples
| Mean (μ)100Std Dev (σ)15k Standard Deviations1RuleEmpirical rule | Interval [μ - kσ, μ + kσ]: [85, 115]Empirical-rule coverage: 68.27%Chebyshev minimum coverage: 0%Selected benchmark: 68.27% |
| 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% |
How it works
The interval runs from μ − kσ to μ + kσ. For a normal distribution, the share inside comes from the normal curve. Chebyshev's inequality gives a lower bound of 1 − 1/k² that applies to any data, which is much more cautious.
The formula
| k | Number of standard deviations |
Practical tips
- With μ = 100 and σ = 15, two standard deviations cover 70 to 130, about 95.45% of a normal distribution.
- Chebyshev guarantees at least 75% within 2 standard deviations, and at least 88.9% within 3.
- Check your data are roughly bell-shaped before relying on the 68-95-99.7 figures.
Frequently asked questions
Why 68-95-99.7?
Those are the rounded shares within 1, 2 and 3 standard deviations of a normal mean.
When should I use Chebyshev?
When you don't know the shape of the distribution, or it's clearly skewed.
Does it work for k = 1.5?
Yes. Any positive k works, and Chebyshev's bound only means something for k above 1.