FIND ME ON

GitHub

LinkedIn

Markov's Inequality

🌱

Theorem
ProbabilityStochasticProcesses

Let ZZ be a integrable random variable. Then for any ϵ>0\epsilon>0, P(Zϵ)E[Z]ϵ\mathbb{P}(Z\ge\epsilon)\le\frac{\mathbb{E}[Z]}{\epsilon}

\begin{proof} Define the event E={Xα}={ωΩ:X(ω)α}E=\{ X\ge\alpha \}=\{ \omega \in\Omega:X(\omega)\ge \alpha \}. Then Xα1EX\ge \alpha \mathbb{1}_{E}where if ωE\omega \in E then {X(ω)αα1E(ω)=α    X(ω)α1E(ω)\begin{cases} X(\omega)\ge \alpha \\ \alpha \mathbb{1}_{E}(\omega)=\alpha \end{cases}\implies X(\omega)\ge \alpha \mathbb{1}_{E}(\omega)and if ω∉E\omega \not\in E then {0X(ω)<αα1E(ω)=0    X(ω)α1E(ω)\begin{cases} 0\le X(\omega)<\alpha \\ \alpha \mathbb{1}_{E}(\omega)=0 \end{cases}\implies X(\omega)\ge \alpha \mathbb{1}_{E}(\omega)hence E[X]αE[1E]    E[X]αP(Xα)\begin{gather*} \mathbb{E}[X]\ge\alpha \mathbb{E}[\mathbb{1}_{E}]\\ \iff\\ \frac{\mathbb{E}[X]}{\alpha}\ge\mathbb{P}(X\ge \alpha) \end{gather*} \end{proof}

Linked from