3
$\begingroup$

There is a property for continuous Markov process that each point $y$ in its state space is hit with positive probability one starting from any interior point $x$.

This property is called the regularity of continuous Markov process. For example, $X_{t}$ is the 1-dimensional brownian motion. The state space is $(-\infty, +\infty)$. I found this concept from the paper: on increasing continuous Markov processes by E.CINLAR. Maybe there is another name from standard text book.

My question is as follows. Suppose $X_{t}$ is a Levy process which is not a pure jump process. This means $\sigma\neq 0 $ in its generating triplet $(\sigma, \gamma, \nu)$. Is $X_{t}$ regular?

Any references are very appreciated.

$\endgroup$
1
$\begingroup$

A counterexample is to let $X_t$ be Brownian motion with drift. Start at any point $x$ and suppose the drift is negative. Let $N_y$ be the event that $y$ is never hit, i.e., $N_y=\{(\forall t)\, X_t < y\}$. With probability one there will be some positive value that is not hit; see e.g. this question. So $$ \mathbb P (\cup_{y\in\mathbb N}\, N_y) = 1. $$ Therefore $$ \exists y\in\mathbb N\qquad \mathbb P(N_y)>0, $$ and such a $y$ is a counterexample to regularity.

$\endgroup$
  • $\begingroup$ Thanks your answer. It's very helpful. Maybe the way I asked or that paper is wrong. It's worthwhile to checking it carefully. $\endgroup$ – Hengyu Zhou Feb 22 '14 at 9:18
  • 1
    $\begingroup$ I do not know where in the paper you found the statement "hit with probability one". I can only find "hit with positive probability". $\endgroup$ – Stephan Sturm Feb 22 '14 at 16:39
  • 1
    $\begingroup$ (I answered the probability-one version of the question, version number 1) $\endgroup$ – Bjørn Kjos-Hanssen Feb 25 '14 at 1:24

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.