• $(\Omega,\mathcal A,\operatorname P)$ be a probability space
  • $I=\mathbb N_0$ or $I=[0,\infty)$
  • $(E,\mathcal E)$ be a measurable space
  • $\mu$ and $\nu$ be probability measures on $(E,\mathcal E)$
  • $X$ be an $(E,\mathcal E)$-valued time-homogeneous Markov chain on $(\Omega,\mathcal A,\operatorname P)$ with transition semigroup $(\kappa_t)_{t\in I}$

Assume $\mu$ is invariant with respect to $(\kappa_t)_{t\in I}$ and $$\left|\mu-\nu\kappa_t\right|\xrightarrow{t\to\infty}0\tag1,$$ where the left-hand side denotes the total variation distance of $\mu$ and $\nu\kappa_t$ and $\nu\kappa_t$ denotes the composition of $\nu$ and $\kappa_t$.

I want to show that there is a version $X^{(\eta)}$ of $X$ with $\operatorname P\left[X^{(\eta)}_0\in\;\cdot\;\right]=\eta$ for $\eta\in\left\{\mu,\nu\right\}$ with $$\tau:=\inf\left\{t\in I:X^{(\mu)}_s=X^{(\nu)}_s\text{ for all }s\in I\text{ with }s\ge t\right\}<\infty\;\;\;\operatorname P\text{-almost surely}.\tag2$$

As pointed out by E-A, the basic idea is the following: For any $t\ge0$, there is (see Theorem 2.12) a probability measure $\eta_t$ (called a coupling of $\mu$ and $\nu\kappa_t$) on $(E\times E,\mathcal E\otimes\mathcal E)$ with $$\eta_t(B\times E)=\mu(B)\text{ and }\eta_t(E\times B)=(\nu\kappa_t)(B)\;\;\;\text{for all }B\in\mathcal E\tag3$$ and $$\left|\mu-\nu\kappa_t\right|=\eta_t(\Delta^c),\tag4$$ where $$\Delta:=\left\{(x,x):x\in E\right\}.$$

Clearly, if we would be able to show (are we?) that there is a $(\mathcal A,\mathcal E\otimes\mathcal E)$-measurable $Y_t:\Omega\to E\times E$ with distribution $\eta_t$ under $\operatorname P$, we would obtain $$\eta_t(\Delta^c)=\operatorname P\left[(Y_t)_1\ne(Y_t)_2\right]\tag5.$$ However, even when we can show this, why can we choose $(Y_t)_{t\ge0}$ such that $X^{(\mu)}:=((Y_t)_1)_{t\ge0}$ and $X^{(\nu)}:=((Y_t)_2)_{t\ge0}$ are versions of $X$?


Let me try this:

Coupling $\rightarrow$ TV goes to 0:

We can argue that $P(\tau > t) \geq P(X_t \not = Y_t)$, since if $X_t$ is not equal to $Y_t$, then $\tau$ has to be greater than t. We also know that $P(X_t \not = Y_t)$ is an upper bound on the TV (or rather, the TV is a lower bound on this quantity), since the $P(X_n \not =Y_n) \geq | \mu(A) - \nu(A) | $ for any $A$. (You can find the proof in a lot of sources; if I were to summarize it, you should write $P(X_t \in A) = P(X_t = Y_t, X_t \in A) + P(X_t \not = Y_t, Y_t \in A)$, and do the same for $Y_t$; the first terms will cancel. http://websites.math.leidenuniv.nl/probability/lecturenotes/CouplingLectures.pdf; you can also graphically observe what the total variation distance is by drawing densities together and you note that it is the half of the not-overlapping area, and that gives you a pictorial proof of this statement, since you know that you can't set $X_t = Y_t$ in any of those areas.)

TV goes to 0 $\rightarrow$ coupling:

Let us construct our coupling in the most natural sense: let $X_t = Y_t$ once $X_{t_0} = Y_{t_0}$ for some $t_0$. We need to start by arguing for $P(X_t \not = Y_t | X_0, Y_0) \leq 1-\epsilon$ for some $t, \epsilon$. Now, we know that we can construct a coupling such that $P(X_t \not = Y_t) = TV(\mu, \nu_t)$ for some fixed $t$. (Proof is in the pdf posted above; Theorem 2.12; you can also argue it from the picture again). Since the TV is going to 0, you can find a $t, \epsilon$ such that the claim is true. Now, you can repeat the same argument and invoke the time homogeneity of your Markov process to show that $P(X_{2t} \not = Y_{2t} | X_t \not = Y_t) \leq 1 - \epsilon$. Now, you can inductively show a geometric decay $P(X_{2t} \not = Y_{2t}) \leq (1 - \epsilon)^k$, which is enough to show that this time is almost surely finite. (Adapted this proof from https://www2.cs.duke.edu/courses/fall15/compsci590.4/slides/lec5.pdf; Section 4, Theorem 1, I might have messed up in certain places)

Feel free to point things out if there is anything wrong/unclear/unjustified. The first one is I think OK (it is a fairly standard proof); the second one involves constructing the coupling, so not sure if all of my steps were completely justified)

  • $\begingroup$ Sorry for the late response. Can you tell me how $\hat X$ and $\hat X'$ in Theorem 2.12 of the first PDF are defined? $\endgroup$ – 0xbadf00d Jan 2 at 0:02
  • $\begingroup$ The first direction is now clear to me. Thank you very much. However, I'm struggling with the other direction. I guess $X$ is supposed to be $X^{(\mu)}$ and your $Y$ is supposed to be $X^{(\eta)}$, right? The first question I have is why $\hat X$ and $\hat X'$ (from my former comment) even exist. The second one is: Why are $X^{(\mu)}$ and $X^{(\nu)}$ versions of $X$? (I've upvoted your answer and will accept it, once we've (hopefully) solved the remaining questions.) $\endgroup$ – 0xbadf00d Jan 3 at 0:12
  • $\begingroup$ Still interested in an answer. Could you take a look at my comments? $\endgroup$ – 0xbadf00d Jan 11 at 18:02
  • $\begingroup$ Yes; feel free to re-label them. Their existence is essentially proof of Theorem 2.12 in that PDF; in your notation $X$ and $\hat{X}$ are the same map from $\Sigma \rightarrow E$. It's just that one of them is defined with the measure $\mu$ and the other one is defined with $\nu$; the underlying space is otherwise the same. This also hopefully answers the second question too (they are fundamentally the same map). Feel free to ask a separate question about this though; I am not sure how technical of an answer I can give. $\endgroup$ – E-A Jan 12 at 21:46

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.