1
$\begingroup$

Information:

a-) $X$ and $Y$ are two continuous random variables on $\mathbb{R}$ having continuous distribution functions $F$ and $G$ with $G(y)\geq F(y)$ for all $y$.

b-) $S^X_n=\sum_{i=1}^n X_i$, $S^Y_n=\sum_{i=1}^n Y_i$, $A>0$, and $B<0$; where $X_i$ and $Y_i$ are i.i.d. replicas of $X$ and $Y$ respectively.

c-) $E[X]<0$ and $E[Y]<0$.

d-) $\partial F/\partial G$, (the ratio of densities) is an increasing function on its domain.

What I know:

By coupling (since $G\geq F$), I know that there exist a pair of random variables $(X^{'},Y^{'})$ such that $X=X^{'}$ in distribution, $Y=Y^{'}$ in distribution, and $X^{'}\geq Y^{'}$ almost surely. Using this result, I also have $S_n^{X^{'}}=\sum_{i=1}^n X^{'}_i\geq \sum_{i=1}^n Y^{'}_i=S_n^{Y^{'}}$. Since this holds for all $n$, I am able to compare the following:

$$\tau_A^{X^{'}}=\inf\{n\geq 0:S_n^{X^{'}}\geq A\}$$ $$\tau_A^{Y^{'}}=\inf\{n\geq 0:S_n^{Y^{'}}\geq A\}$$ $$\tau_B^{X^{'}}=\inf\{n\geq 0:S_n^{X^{'}}\leq B\}$$ $$\tau_B^{Y^{'}}=\inf\{n\geq 0:S_n^{Y^{'}}\leq B\}$$

with $\tau_A^{X^{'}}\leq \tau_A^{Y^{'}}$ since $S_n^{Y^{'}}\geq A$ implies $S_n^{X^{'}}\geq A$ and similarly $\tau_B^{X^{'}}\geq \tau_B^{Y^{'}}$. Please see (for details).

Question: I wonder if

$$E[\min\{\tau_A^{X^{'}},\tau_B^{X^{'}}\}]\geq E[\min\{\tau_A^{Y^{'}},\tau_B^{Y^{'}}\}]$$

holds for any $(A,B)$ and $(X,Y)$.

Added:(12.05.2014)

I thought about using $$\min(a,b)=\frac{1}{2}(|a+b|-|a-b|)$$ the inequality became $$(E[\tau_A^{X^{'}}]-E[\tau_B^{Y^{'}}])-(E[\tau_A^{Y^{'}}]-E[\tau_B^{X^{'}}])\geq E[|\tau_A^{X^{'}}-\tau_B^{X^{'}}|]-E[|\tau_A^{Y^{'}}-\tau_B^{Y^{'}}|]$$ But I know $E[\tau_A^{X^{'}}]=\infty$ and therefore I dont know how one could go forward and what would be the role of increasing $\partial F/\partial G$.

Added:(14.05.2014)

Please correct me if I am mistaken. It seems that we also have $$E[\min\{\tau_A^{X^{'}},\tau_B^{X^{'}}\}]=E[|\tau_A^{X^{'}}-\tau_B^{X^{'}}|]$$ $$E[\min\{\tau_A^{Y^{'}},\tau_B^{Y^{'}}\}]=E[|\tau_A^{Y^{'}}-\tau_B^{Y^{'}}|]$$

writing this result above we see that

$$E[\tau_B^{X^{'}}]-E[\tau_B^{Y^{'}}] \geq E[\tau_A^{Y^{'}}]-E[\tau_A^{X^{'}}]\quad\quad\quad\quad (*)$$

implies

$$E[\min\{\tau_A^{X^{'}},\tau_B^{X^{'}}\}]\geq E[\min\{\tau_A^{Y^{'}},\tau_B^{Y^{'}}\}]$$

therefore to show that $(*)$ indeed holds will be enough. Once again $E[\tau_A^{X^{'}}]=\infty$ and $E[\tau_A^{Y^{'}}]=\infty$. If such a comparison is possible then I would guess that $E[\tau_A^{Y^{'}}]-E[\tau_A^{X^{'}}]\geq 0$ because of stochastical ordering. We also know $E[\tau_B^{X^{'}}]-E[\tau_B^{Y^{'}}]\geq 0$.

My idea was using $\tau_A^{X^{'}}=\tau_{A_{\lim_{B\rightarrow\infty}}}^{X^{'}}=\inf\{n\geq 0:S_n^{X^{'}}\geq A,S_n^{X^{'}}\leq B\}$ and the same for $\tau_A^{Y^{'}}$. I hope that this formulation could give me $$E[\tau_{A_{\lim_{B\rightarrow\infty}}}^{Y^{'}}]-E[\tau_{A_{\lim_{B\rightarrow\infty}}}^{X^{'}}]=E[\tau_A^{Y^{'}}]-E[\tau_A^{X^{'}}]<\infty$$

Can?

I asked the same question in mathstackexchange in two parts. I have a counterexample if $\partial F/\partial G$ is not necessarily increasing. However, I wonder very much if the claim holds with this monotonicity condition or not. Unfortunately, I seem to be hopeless to deal with this problem.

I will really appreciate any help. Thank you very much.

$\endgroup$
3
  • $\begingroup$ I do not understand the notation in d), but I'll take it to mean that the ratio of densities is increasing. Here is what a counterexample would look like. I did not work through the details. Let f be a density concentrated just to the right of A. Let g be a density concentrated around B/N for some large N, with a small mass eps at M*B, for some large M and small eps. Let X have unnormalized density f(x)+g(x)/sqrt(1-x). Let Y have unnormalized density g(x)+f(x)/sqrt(1+x). Add in eps times a normal(0,1) if you want fully support. Then normalize. (continued...) $\endgroup$
    – guest
    May 12, 2014 at 22:50
  • $\begingroup$ With a suitable choice of constants (sufficiently big M) the conditions will be satisfied. Y will take O(N) steps to cross the lower boundary with high probability, while X will cross the upper boundary in O(1) steps with high probability. The expectations should behave similarly. Again, this isn't a complete answer, but I hope it is a help. $\endgroup$
    – guest
    May 12, 2014 at 22:50
  • $\begingroup$ How can $f$ be concentrated to the right of $A$? you mean the expected value? if so it cannot be. $\endgroup$ May 12, 2014 at 23:15

1 Answer 1

0
$\begingroup$

I found a counter example for the claim with mean shifted Gaussian distributions. So the claim is not true even with the condition in $d$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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