k = 2
1/k² = 0.25
At least 75% within 2σ.
For any random variable with mean and standard deviation , . At least of the distribution lies within standard deviations of the mean.
This page is easier if you are comfortable with:
For the probability of being 2 or more standard deviations away is at most ; for , at most , so at least 88.9% lies within . If , : . To guarantee a tail probability of at most 0.04 we need . The bound holds for every distribution, which is why it is conservative: for a normal variable the actual probability beyond is 0.0455 against the bound 0.25 (standardising a normal variable).
The proof applies Markov's inequality to , whose expectation is the variance (variance and standard deviation). The inequality is the tool for proving the weak law of large numbers: applied to the sample mean (variance ) it shows .
Sharper bounds need more assumptions: for bounded or sub-Gaussian variables see Chernoff and Hoeffding bounds; for the normal case use exact tables (the central limit theorem).
Chebyshev's inequality bounds how likely a value is to be far from the mean, whatever the distribution.
The bound falls quickly as k increases.
These examples are worked by hand, step by step, so you can compare your own working.
Bound .
If , , bound .
What guarantees a tail probability of at most 0.04?
A few situations come up again and again. Knowing them saves time and helps you spot answers that cannot be right.
1/k² = 0.25
At least 75% within 2σ.
1/9
At least 88.9% within 3σ.
k = 3, P ≤ 1/9
Bound 0.1111.
40 to 60
μ ± 2σ.
P(|X − μ| ≥ kσ) ≤ 1/k² for any distribution with finite variance, so at least 1 − 1/k² of the probability is within k standard deviations.
Find the Chebyshev bound for k = 2.
Find the bound for k = 3.
μ = 50, σ = 5. Bound P(|X − 50| ≥ 15).
Lower end of the 75% interval μ − 2σ.
bound 1/2 ✗
1/k² ✓
Squared.
P = 1/k² ✗
upper bound only ✓
Often very loose.
meaningful ✗
bound exceeds 1 ✓
Useless.
only normal ✗
any distribution ✓
Distribution-free.
Try each one on paper, then open the answer. Use the solver above to see the full working.
0.0625
0.04
0.0625
0.8889
0.75
116
Unlimited fresh questions with instant marking. Choose a level, work it out on paper, then type your answer.
Solve: …
P(|X − μ| ≥ kσ) ≤ 1/k².
At least 75% within 2σ.
Often not.
No.
Markov’s inequality.
central limit theorem.
More in Advanced statistics and stochastic processes and the formula sheets.