Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

$X$ is a positive continuous random variable. $E[X^p]$ is the $p$-th moment of $X$, $p\ge2$. Is the following moment inequality valid? $E[X^p]\le (p-1)^{p/2}(E[X^2])^{p/2}$

If so, What is the name of this inequality, and how to prove it?

share|cite|improve this question
Try $p=1$. (extra characters) – Byron Schmuland Mar 20 '12 at 15:36
Have you tried playing around with $E(X^p)=\int_0^\infty py^{p-1}P(X>y)dy$, which holds for $p>0$? As Byron mentions, it's not true for $p=1$, but maybe try looking at $1<p<2$ and that $P(X>y)dy/E(X)$ is a probability measure along with Jenson's Inequality. – Alex R. Mar 20 '12 at 17:25
Let me up the ante:) For $1<p<2$ consider the degenerate random variable $X\equiv 1$. – Byron Schmuland Mar 20 '12 at 20:35
Sam, could you please explain your answer in more detail? I just modify p>=2 and p is integer. I really appreciate your answers! – huoshanzhx Mar 24 '12 at 2:59

No inequality $\mathrm E(X^p)^2\leqslant c(p)\cdot\mathrm E(X^2)^p$ $(\ast)$ may hold for a finite $c(p)$ and for every nonnegative random variable $X$, as soon as $p\gt2$.

To see why, note that the LHS of $(\ast)$ may be infinite while its RHS is finite. This happens, for example, when the tail distribution is $\mathrm P(X\geqslant x)\sim c/x^q$ when $x\to+\infty$, for some $c\gt0$ and $2\lt q\leqslant p$.

The same argument also shows that, even when restricted to bounded random variables (and in particular, with every moment finite), $(\ast)$ cannot hold for $p\gt2$, for any finite $c(p)$. To wit, consider a random variable $X$ as above and, for every $u\gt0$, $X_u=\min\{X,u\}$. Then every moment of every $X_u$ is finite and, when $u\to+\infty$, $\mathrm E(X_u^2)\to\mathrm E(X^2)$, which is finite, and $\mathrm E(X_u^p)\to+\infty$. Hence, for every finite $c$, one sees that $\mathrm E(X_u^p)^2\gt c\cdot\mathrm E(X_u^2)^p$ for some finite $u\gt0$, and in particular for some bounded random variable $X_u$.

If $1\leqslant p\lt2$, any nonzero $X$ which is almost surely constant disproves the proposed inequality. Finally, if $p=2$, the proposed inequality reduces to a tautology.

share|cite|improve this answer
Thanks, Didier Piau. What about p>=2 and p is integer? – huoshanzhx Mar 24 '12 at 3:03
I am an engineering person. Usually, for our problems, the first 5 moments are finite – huoshanzhx Mar 24 '12 at 3:06
The case p>2 and p integer is no different from the case p>2. // See Edit for some explicit counterexamples with every moment finite. – Did Mar 24 '12 at 10:10

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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