I know that Cantelli's Inequality states that for a random variable $X$ with mean $μ$ and variance $\sigma^2$:

$$ P(X-μ\geq \alpha)\leq\frac{\sigma^2}{\sigma^2 + \alpha^2} $$

Here, I am trying to figure out how to show that:

$$ P(|X-μ| \geq \alpha)\leq\frac{2\sigma^2}{\sigma^2 + \alpha^2} $$

Is there an easy way to do this directly from Cantelli's inequality or would it require a separate proof? thanks.


Apply your known formula (Cantelli's inequality) twice, once on $X$ and once on $Y=-X$ to get the result. The random variable $Y=-X$ has mean $μ_Y=Ε[Y]=-Ε[Χ]=-μ$ and variance $σ_Y=Var(Y)=(-1)^2Var(X)=σ^2$. So \begin{align}P(|X-μ|\ge α)&=P(X-μ\ge α)+P(X-μ\le -α)\\[0.2cm]&=P(X-μ \ge α)+P(-X-(-μ)\ge α)\\[0.2cm]&=P(X-μ\ge α)+P(Y-μ_Y\ge α)\end{align}

  • $\begingroup$ Hi, thanks for the response. I am just beginning to take probability so sorry if this sounds stupid but what allows us to get the first step above, $P(|X-μ|\ge α)=P(X-μ\ge α)+P(Χ-μ\le -α)$? The farthest I got was that $P(|X-μ|\ge α) = P(X-μ\ge α \lor Χ-μ\le -α)$ but dont know why you can add the inside, since I thought you could only add events when they are disjoint. How do I know these events are disjoint so to say? Thank you so much! $\endgroup$ – user136503 Jan 29 '16 at 15:03
  • $\begingroup$ $P(A\cup B)=P(A)+P(B)$ when the events $A$ and $B$ are disjoint. You certainly have had this formula before going to random variables? So, here the events $X-μ\ge α$ and $Χ-μ\le -α$ are disjoint, hence the probability of their union is the sum of their probabilities. $\endgroup$ – Jimmy R. Jan 29 '16 at 15:08
  • $\begingroup$ Thanks, so the two events are disjoint because they don't overlap? Is there a way to formally think about it instead of drawing number lines? (which is what I do) $\endgroup$ – user136503 Jan 29 '16 at 15:10
  • $\begingroup$ Yes, there are ways with $X(ω)$ and so on, but drawing number lines is also the way I think about it. $\endgroup$ – Jimmy R. Jan 29 '16 at 15:16
  • 1
    $\begingroup$ Awesome, thank you! $\endgroup$ – user136503 Jan 29 '16 at 15:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.