1
$\begingroup$

I have a problem for which I either need a proof or a counterexample.

We are given two discrete random variables $x_1$ and $x_2$ in $[0, n]$ where $F_1(x)$ is the probability of $x_1\leq x$, and similarly $F_2(x)$ is the probability of $x_2 \leq x$. My goal is to show that there always exists a number $c > 0$ that satisfies $$\int_{0}^{\infty}x{f_2(x)}(1-F_1(x))\leq c\int_{c}^{\infty}f_1(x)F_2(x),$$ where $f_1(x)= \Pr[x_1=x]$ and $f_2(x)=\Pr[x_2=x]$.

Thanks for any help.

$\endgroup$
0
$\begingroup$

The OP has just been edited from an apparently absolutely continuous version to a discrete one. So, I gather that the integrals are now to be understood with respect to the counting measure, that is, as the sums.

However, there is no essential difference here between the absolutely continuous and discrete versions, and counterexamples for the two versions can be quite similar. For instance, suppose that $x_1$ and $x_2$ are random variables taking only nonnegative integral values such that \begin{equation} F_1(x)=1-\frac1{\ln x}\quad\text{and}\quad f_2(x)=\frac C{x\ln^2 x} \end{equation} for integers $x\ge3$ and $F_1(x)=f_2(x)=0$ for integers $x<3$, where $C$ is a positive real constant such that $f_2$ is a pmf. Then \begin{multline} \int_{0}^{\infty}x{f_2(x)}(1-F_1(x))=\sum_{0}^{\infty}x{f_2(x)}(1-F_1(x)) =\sum_{3}^{\infty}\frac C{\ln^3 x}=\infty \\ >c\ge c\int_{c}^{\infty}f_1(x) \ge c\int_{c}^{\infty}f_1(x)F_2(x) \end{multline} for any real $c>0$, so that we have the opposite of the desired inequality.

$\endgroup$
1
$\begingroup$

The functions $F_1(x) = 1-\frac{1}{(x+1)^3}$, $F_2(x) = 1-e^{-x}$ (so that $f_1(x) = \frac3{(x+1)^4}$, $f_2(x) = e^{-x}$) seem to provide a counterexample.

Any context for this question?

$\endgroup$

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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