E(X) = 10, a = 50
P(X ≥ 50) ≤ 0.2
Bound.
For a non-negative random variable and , .
This page is easier if you are comfortable with:
If the average is 10, at most one in five values can be 50 or more: . If mean income is 40 (thousand), at most earn 200 or more. If the mean waiting time is 5 minutes, . For an exponential waiting time the exact probability is , so the bound is far from tight, but it needs only the mean (the exponential distribution).
The proof is one line: , because the part of the expectation from values at least is already at least times their probability (expectation of a discrete variable). Applying it to gives Chebyshev's inequality, and applying it to gives the Chernoff bounds (the moment generating function, Chernoff and Hoeffding bounds).
In combinatorics the case is the first moment method: . The requirement is essential; for variables that can be negative use a shifted or squared version.
Markov's inequality limits how often a non-negative quantity can be many times larger than its mean.
The chance of reaching a threshold falls as 1/a.
These examples are worked by hand, step by step, so you can compare your own working.
Bound when .
Bound the proportion earning at least 200 when the mean is 40.
Bound when the mean is 5 minutes.
A few situations come up again and again. Knowing them saves time and helps you spot answers that cannot be right.
P(X ≥ 50) ≤ 0.2
Bound.
0.25
Bound.
bound 0.5
At most half exceed twice the mean.
apply to (X − μ)²
k = 2: 0.25.
For X ≥ 0 and a > 0, P(X ≥ a) ≤ E(X)/a. It needs only the mean, so it is weak but universal.
E(X) = 10. Bound P(X ≥ 50).
E(X) = 3. Bound P(X ≥ 12).
Bound P(X ≥ 2E(X)).
Apply to (X − μ)² with a = 4σ². Find 1/4.
applying to X < 0 ✗
needs X ≥ 0 ✓
Non-negativity.
meaningful ✗
trivial if a < E(X) ✓
Check a.
P = E/a ✗
upper bound ✓
Usually loose.
a/E ✗
E/a ✓
Check.
Try each one on paper, then open the answer. Use the solver above to see the full working.
0.2
0.2
0.25
0.25
0.25
0.3333
Unlimited fresh questions with instant marking. Choose a level, work it out on paper, then type your answer.
Solve: …
P(X ≥ a) ≤ E(X)/a.
X ≥ 0 and a > 0.
Rarely.
Apply Markov to (X − μ)².
Chebyshev’s inequality.
More in Advanced statistics and stochastic processes and the formula sheets.