TOPICS
Search

Markov's Inequality


Markov's inequality bounds the probability that a nonnegative random variable x reaches or exceeds a positive threshold. For every real number a>0,

 P(x>=a)<=(<x>)/a.
(1)

Here <x> denotes the expectation value of x. The bound also holds when <x>=infty, although it then gives no restriction on the probability. For the indicator function 1_A of the event A={x>=a}, the pointwise inequality a1_A<=x gives aP(A)<=<x>. Dividing by a>0 proves the bound for arbitrary probability distributions (Durrett 2019, pp. 29-30).

When x has a probability density function P(x), the theorem also has an integral proof. Write

<x>=int_0^inftyxP(x)dx
(2)
=int_0^axP(x)dx+int_a^inftyxP(x)dx.
(3)

Since the probability density function P(x) and x are nonnegative,

<x>=int_0^axP(x)dx+int_a^inftyxP(x)dx
(4)
>=int_a^inftyxP(x)dx
(5)
>=int_a^inftyaP(x)dx
(6)
=aint_a^inftyP(x)dx
(7)
=aP(x>=a).
(8)

Q.E.D.


See also

Chebyshev Inequality, Chernoff Bound, Expectation Value

Explore with Wolfram|Alpha

References

Durrett, R. Probability: Theory and Examples. Version 5, Jan. 11, 2019. Author manuscript. https://services.math.duke.edu/~rtd/PTE/PTE5_011119.pdf.

Referenced on Wolfram|Alpha

Markov's Inequality

Cite this as:

Weisstein, Eric W. "Markov's Inequality." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/MarkovsInequality.html

Subject classifications