I have the following, seemingly simple question:

Consider a stochastic process $(X_t)$ satisfying $X_t\le X_s$ a.s. for all $t\le s.$ My question is: Does there exist a modification $\tilde{X}$ of $X$, which almost surely has increasing sample paths $t\mapsto\tilde{X}_t(\omega)$?

I assume such a modification exists, but I did not manage to proof it.

Thanks in advance!

  • 2
    $\begingroup$ Does $\tilde X_t(\omega)=\mathop{ess\ sup}_{s\le t}X_s(\omega)$ work? $\endgroup$ – Anthony Quas Jun 1 '14 at 21:11
  • $\begingroup$ How do you show measurability? However the idea seems to work (sup over rational times, countability of times with essential jumps, etc.) $\endgroup$ – fedja Jun 2 '14 at 1:33

Yes, this works. As suggested by fedja, focus for a moment on the rational times. The function (on $\mathbb Q$) $f_{\omega}(t)=X_t(\omega)$ is almost surely increasing. Set $\widetilde{X_t}=0$ for all $t\in\mathbb R$ on the exceptional set. Now if $t\in\mathbb R$ is arbitrary, then $f_{\omega}(t-)\le X_t(\omega) \le f_{\omega}(t+)$ a.s. because only countably many times are involved (of course, the null set will depend on $t$). Set $\widetilde{X_t}(\omega)=f_{\omega}(t+)$ (say) on the null set where this fails.

All paths are increasing now because the required inequalities hold if at least one rational time is involved.


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.