4
$\begingroup$

Any ideas on how I could prove the veracity or falseness of the following inequality?

Let $X:\Omega \to \mathbb{R}$ a random variable such that the expressions under are well-defined. Then

$$E[e^X] \leq 1 + e^{E[|X|]}.$$

I have the feeling that this is true but i do not know how to show it. I was thinking Jensen's inequality but it goes the wrong way.

One more question, if the above inequality is false, is there a way to upperbound $E[e^X] \leq f(E|X|)$ ?

Thanks you very much for your help.

$\endgroup$
2
$\begingroup$

The inequality is not true: take $X$ with $P\{X=n\}=\frac 1n$ and $P\{X=0\}=1-\frac 1n$. Then the LHS is $\frac{e^n}n+1-\frac 1n$ and the RHS is $1+e$ (or $f(1)$ in the general version).

$\endgroup$
  • $\begingroup$ But if I apply Jensen's inequality to $t\mapsto e^t$ (since it is convex) I obtain $e^{E[X]} \leq E[e^X]$ and I would like some how, the opposite bound. Nevertheless, it seems that $E[e^X]\leq e^{E[|X|]}$ is true? $\endgroup$ – marc Aug 27 '13 at 15:32
  • $\begingroup$ I've edited, since everything I wrote was misleading. $\endgroup$ – Davide Giraudo Aug 27 '13 at 15:49
  • $\begingroup$ @DavideGiraudo: Just a nit-pick, but since $\mathrm{E}[X]=1$, isn't $1+e^{\mathrm{E}[X]}=1+e$, not $2$? $\endgroup$ – robjohn Aug 27 '13 at 15:56
  • $\begingroup$ Oh thanks! Yes I understood, I also mix Jensen's inequality often. Do you think there is any chance to bound $E[e^X]$ by means of $e^{E[|X|]}$?, some function :) $\endgroup$ – marc Aug 27 '13 at 15:58
  • 1
    $\begingroup$ Sure I see the point.. Thanks a lot :) $\endgroup$ – marc Aug 27 '13 at 16:09

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.