0
$\begingroup$

This is a cross-post of this and this questions from math.stackexchange.com since I have not received any response there. I would like to seek help here.

Suppose $x(t,\omega): [0,T]\times\Omega\rightarrow \mathbf R$ is a random variable on a probability space $\Omega$.

  1. Sequence $(t_k<2^{-k})_{k=1}^\infty \implies \lim_\limits{i\rightarrow\infty}x(t_i,\omega)\rightarrow0$ for almost all $\omega\in\Omega$. Does this imply $\lim_\limits{t\rightarrow0}x(t,\omega)=0$ for almost all $\omega\in\Omega$? What if we assume $\lim_\limits{t\rightarrow0}x(t,\omega)=0$ in probability?

  2. For every sequence $(t_k)_{k=1}^\infty\rightarrow 0$, $\exists$ a subsequence $(t_{k_i})_{i=1}^\infty$ independent of $\omega$ $\ni\big(\lim_\limits{i\rightarrow\infty}x(t_i,\omega)\rightarrow0$ for almost all $\omega\in\Omega\big)$. Does this imply $\lim_\limits{t\rightarrow0}x(t,\omega)=0$ for almost all $\omega\in\Omega$? What if we assume $\lim_\limits{t\rightarrow0}x(t,\omega)=0$ in probability?

$\endgroup$
4
  • $\begingroup$ This is very closely related to your questions on math.stackexchange.com: math.stackexchange.com/questions/2801027/… and math.stackexchange.com/questions/2800921/… $\endgroup$ May 29, 2018 at 22:13
  • $\begingroup$ @AnthonyQuas: Yes it is. I hope I have more luck getting an answer here as I have not received any response there. Do you have any idea regarding the questions? $\endgroup$
    – Hans
    May 29, 2018 at 22:20
  • 1
    $\begingroup$ This isn't cool. You only posted there 4 hours ago. At the very least, you should indicate on your post that this is a cross-posting. Also, you should wait longer before cross-posting. $\endgroup$ May 29, 2018 at 22:21
  • $\begingroup$ @AnthonyQuas: I have now indicated as much. $\endgroup$
    – Hans
    May 29, 2018 at 22:28

1 Answer 1

3
$\begingroup$

No. A counterexample for all of your questions is as follows. Let $\Omega$ be $[0,1)$, with probability measure $\mathbb P$ being Lebesgue measure. Set $x(t,\omega)=1$ if the fractional part of $1/t$ is $\omega$ and 0 otherwise.

This is a version of the standard example satisfying convergence in probability, but not pointwise.

$\endgroup$
3
  • $\begingroup$ Since you clearly indicated a preference to avoid rapid-fire cross-posting, maybe it would be better to move this counterexample to the MSE question(s) you found? $\endgroup$
    – LSpice
    May 29, 2018 at 22:54
  • $\begingroup$ Thank you very much, Anthony. +1 If you want to do as @LSpice says, I have no objection. Just let me know your preference. $\endgroup$
    – Hans
    May 29, 2018 at 23:25
  • $\begingroup$ Anthony, this question is actually posed in service of my other question mathoverflow.net/q/300557/32660. I have proved the convergence in probability mathoverflow.net/a/300960/32660 but am not able to prove the point-wise convergence. Would you be interested in taking a look at it? Thank you. $\endgroup$
    – Hans
    May 29, 2018 at 23:40

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.