0
$\begingroup$

I'm having trouble proving the following statement: let $\Omega \subset \mathbb{R}^n$ be an open set, let $h: \Omega \mapsto \mathbb{R}$ and $\rho : \Omega \mapsto [0,+\infty)$ be Lebesgue-measurable functions and suppose that $\int_{\Omega} \rho \, d \mu =1$. Prove that, if $\mu$ denotes the Lebesgue measure on $\mathbb{R}^n$, for every $p \in [1,+\infty)$: $$ \left( \int_{\Omega} |h| \rho \, d \mu \right)^p \le \int_{\Omega} |h|^p \rho \, d \mu $$ At first sight, I was thinking about Jensen's inequality because $f(x)=|x|^p$ is convex if $p \ge 1 $, but the hypothesis $\mu(\Omega)=1$ is not satified, so we cannot apply that, at least not directly. On the other hand, I can't figure out how classical measure theory inequalities (for instance, Holder's) can work here. I would appreciate any hint.

$\endgroup$
0
$\begingroup$

This is Jensen's inequality -- let $d \nu = \rho d \mu$. Then, its easy to see $\nu(\Omega) = 1$ and $\nu$ is thus a probability measure, and you have the usual Jensen's inequality with respect to $\nu$.

If you want to think of it in probability terms, $\rho$ is a probability density function by virtue of its definition wrt the lebesgue measure. The left hand side is $E[|h(X)|]^p$ and the right is $E[|h(X)|^p]$ where $X \sim \rho$.

$\endgroup$
  • $\begingroup$ Thank you very much! I think I was missing the passage $\int_{\Omega} |h| \, d \nu = \int_{\Omega} |h| \frac{d \nu}{d \mu} \, d \mu = \int_{\Omega} |h| \rho \, d \mu$. $\endgroup$ – GaC Jan 11 '17 at 14:07

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.