2
$\begingroup$

I'm looking for an upper bound on $E_X\left[\left\vert X^2-(E[X])^2\right\vert\right]$, where $X$ is a nonnegative random variable, with known $E[X]$ and $E[X^2]$, but unknown higher moments.

A lower bound is $E[X^2]-(E[X])^2$, and I wonder if a there is a constant multiple of this as an upper bound?

Improve this question
$\endgroup$
4
  • 1
    $\begingroup$ Upper bound in terms of what, exactly? The absolute value sign is superfluous, because $E[X^2] \ge E[X]^2$ in any event, so you are asking for an upper bound for the variance, given that you know the variance. $\endgroup$
    – whuber ♦
    Oct 4 at 2:41
  • $\begingroup$ The issue is that the absolute value is a nonlinear function. $\endgroup$
    – axk
    Oct 4 at 2:52
  • $\begingroup$ @Whuber - the absolute value sign is not superfluous: for non-constant random variables, there is a positive probability that $X<E[X]$ in which case $X^2 -(E[X])^2<0$ $\endgroup$
    – Henry
    Oct 4 at 8:48
  • $\begingroup$ @Henry Thank you -- I had misread that by distributing expectation over the absolute value! $\endgroup$
    – whuber ♦
    Oct 4 at 13:14

1 Answer 1

Reset to default
3
$\begingroup$

I would have thought you could say:

$E\big|X^2-(E[X])^2\big| \le E\big|X^2+(E[X])^2\big| = E\big[X^2\big]+(E[X])^2$

As an example to show it can come close:

  • suppose $X=1000$ with probability $\frac1{100}$ and $X=0$ otherwise,

  • so $E[X]=10$ and $E[X^2]=10000$,

  • we get $E[X^2]-(E[X])^2=9900$ and $E[X^2]+(E[X])^2=10100$

  • with $E\big|X^2-(E[X])^2\big|=10098$ closer to the upper bound.

As to your question of a constant multiple of the variance, the answer seems to be no. Adjusting the previous example, suppose $X=1000$ with probability $\frac1{100}$ and $X=999$ otherwise, so $E[X]=999.01$ and $E[X^2]=998020.99$, you have $E[X^2]-(E[X])^2=0.0099$ and $E\big|X^2-(E[X])^2\big|=39.570498$, almost four thousand times as big. This is not the worst case.

Improve this answer
$\endgroup$
3
  • $\begingroup$ You would have thought but...? Or do you think it still? $\endgroup$
    – Richard Hardy
    Oct 4 at 10:30
  • $\begingroup$ @RichardHardy It was a quick response giving a bound which was not of the form requested. I often make mistakes when looking at something new (here I initially persuaded myself $E[X^2]$ was a bound and then found a counterexample), so I used a weaker tense and mood to reflect this. I still think it correct, but I am happy to leave my original words. $\endgroup$
    – Henry
    Oct 4 at 10:37
  • $\begingroup$ Thanks, I understand. $\endgroup$
    – Richard Hardy
    Oct 4 at 10:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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