Set inclusion between space of probabilities

23 Views Asked by At

I'm working with sets of probability measures of the form: $$ \mathscr{P}_p(\Omega) := \left\{\mu \in \mathscr P: \int_\Omega |x|^p d\mu < + \infty\right\} $$ and I'm trying to show that if $p<q$, then $\mathscr P_p(\Omega)\subset \mathscr P_q(\Omega)$

My immediate thought was to try something like Holder's inequality starting with $$ \int|x|^qd\mu = \| |x|^q\cdot 1\|_{L^1(\Omega, \mu)} $$ and trying to find $\|x\|_p$ in some form as an upper bound, but the conjugate exponents don't seem to work out since $p<q$.

My other idea was to consider the largest natural number $m\ge2$ such that $q-mp\ge 0$, and split the integral $$ \int|x|^qd\mu = \int |x|^{mp}|x|^{q-mp}d\mu $$ and try to use a generlized Holder inequality and the fact that $q-mp < p$ since $m$ is maximal. The issue I seemed to run into again was that the conjugate exponent would bring me back to the $q$-norm, which certainly isn't what I want.

I considered Jensen's inequality (which I don't know very well), but I don't think it works out since $t\mapsto t^{p/q}$ isn't convex.

I'd like to still solve this somewhat on my own, so if you can just nudge me in the right direction or critique my ideas, I'd greatly appreciate that. Thank you!

1

There are 1 best solutions below

0
On BEST ANSWER

Counter-example: On $(0,1)$ with Lebesgue measure let $x(t)=\frac 1 {\sqrt t}$. Take $p=1, q=2$.

What is true is $\mathscr P_q(\Omega)\subset \mathscr P_p(\Omega)$ and this follows by Hölder's inequality with exponents $\frac q p$ and $\frac q {q-p}$