If X is nonnegative then Pr(X≥a)≤aE[X].
Proof 1
E[X]=∑xxPr(X=x)≥∑x≥axPr(X=x)
≥∑x≥aaPr(X=x)=aPr(X≥a)
Proof 2
Let I be an indicator for the event X≥a. Then aI≤X because
if X≥a, then a⋅1≤X
if X<a, then a⋅0≤X
Taking the expectations of both sides gives us aE[I]≤E[X], or aPr(X≥a)≤E[X].
Chebyshev’s Inequality
For any random variable X, let Y=(X−E[X])2. Since Y is nonnegative we can apply the Markov Bound to Y, so Pr(Y≥a)≤aE[Y], or similarly Pr(Y≥a2)≤a2E[Y].
Markov Bound
Proof 1
Proof 2
Chebyshev’s Inequality
Estimators
Iterated Expectation