1
$\begingroup$

I have the following problem:

Consider a state space $E$ and a Markov chain $X$ on $E$ with transition matrix $Q$ such that for every $x \in E$, $Q(x,x)<1$. Define: $\tau:=\inf\{n\geq 1:X_n\neq X_0\}$.

  1. Show that $\tau$ is a stopping time
  2. Show that, for every $x\in E$, $\tau<\infty$ $P_x -a.s.$
  3. Compute the distribution of $\tau$ and that of $X_\tau$ under $P_x$

So for the first point I tried this: $\tau$ is a stopping time if $\{\tau =n\}$ is $\mathbb{F}_n$-measurable $\forall n$. Since the Markov chain is based on its natural filtration we know that $\forall n : X_n$ is $\mathbb{F}_n$-measurable.

Define $A:=\{X_n:X_n\neq X_0\}$ and so $\tau = \inf\{n\geq 1: X_n \in A\}$ so I can write $\{\tau =n\}= \{X_n \in A\}\bigcap \{X_{n-1} \not\in A\}\bigcap...\bigcap \{X_0 \not\in A\}$ And by this equation we see that $\{\tau =n\}$ is $\mathbb{F}_n$-measurable $\forall n$.

For points 2 and 3 I can't get the meaning of $P_x$-a.s. I think it should be easy since by assumption I have $ Q(x,x)<1$ (which it should implies that there exist an y s.t. $Q(x,y)>0$. But this is only a thought..

$\endgroup$
  • $\begingroup$ The definition of A seems unsound. You might want to expand it. At the moment, A is a random subset of the state space, is this what you mean and, if so, what is A exactly? $\endgroup$ – Did Nov 24 '13 at 10:14
  • $\begingroup$ Yes, I want A to be a subset of my statespace and the complement of $X_0$ does it make sense? $\endgroup$ – sky90 Nov 24 '13 at 11:10
  • $\begingroup$ Then define $A=E\setminus\{X_0\}$, this is clearer. Indeed, with this definition, $[X_n\in A]$ is in $F_n$ and you are done. To show that $\tau\lt\infty$ almost surely conditionally on $X_0=x$, one may compute $P[\tau=n\mid X_0=x]$ for every $n$, then check that these probabilities sum to $1$. $\endgroup$ – Did Nov 24 '13 at 11:29
  • $\begingroup$ I don't understand really what do you mean with this sum. Could you say me if this other way make sense? I assume that $\tau = \infty$ Then I can say that $\forall y\in E: P_x(H_y<\infty)=0$ where $H_y=\{n\geq 1: X_n =y\}$ so I can say that X is not irreducible (all states do not communicate) but that's a contradiction since $Q(x,x)<1$ $\endgroup$ – sky90 Nov 24 '13 at 12:32
  • $\begingroup$ And for point 3.. What is mean to be the distribution of a stopping time? $\endgroup$ – sky90 Nov 25 '13 at 12:22
1
$\begingroup$

From the looong string of comments above, it seems a problem of understanding might be related to the words distribution of a stopping time. Let us explain these. To this end, one should consider a probability space $(\Omega,\mathcal F,P)$, a filtration $(\mathcal F_n)_{n\in\mathbb N}$ on $(\Omega,\mathcal F)$, and a random variable $S:(\Omega,\mathcal F)\to \mathbb N\cup\{+\infty\}$.

Recall that $S$ has a distribution. This is the image of $P$ by $S$, that is, the measure $\mu$ defined by $\mu(M)=P[S\in M]$ for every $M\subseteq\mathbb N\cup\{+\infty\}$. The measure $\mu$ is entirely characterized by the collection of probabilities $\{\mu(\{n\})\mid n\in\mathbb N\}$, or, equivalently, by the collection of probabilities $\{\mu(\{1,2,\ldots,n\})\mid n\in\mathbb N\}$, or, equivalently, by the collection of probabilities $\{\mu(\{n,n+1,\ldots\})\mid n\in\mathbb N\}$.

Furthermore, $S$ is a stopping time with respect to $(\mathcal F_n)_{n\in\mathbb N}$ if and only if the event $\{S\leqslant n\}$ is in $\mathcal F_n$, for every $n$ in $\mathbb N$. Equivalently, one can ask that the event $\{S=n\}$ is in $\mathcal F_n$, for every $n$ in $\mathbb N$.

Consequently, in the present case, question 1. asks to show that $\{\tau=n\}$ is in $\mathcal F_n$, for every $n$ in $\mathbb N$, for some unspecified (but canonical) filtration $(\mathcal F_n)_{n\in\mathbb N}$ I will let you explain, and the first part of question 3. asks to compute $P[\tau=n]$ for every $n$ in $\mathbb N$, or equivalently, $P[\tau\leqslant n]$ for every $n$ in $\mathbb N$, or equivalently, $P[\tau\geqslant n]$ for every $n$ in $\mathbb N$.

$\endgroup$
  • $\begingroup$ Ok thanks for your help $\endgroup$ – sky90 Dec 4 '13 at 5:18

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.