• $(\Omega,\mathcal A,\operatorname P)$ be a probability space
  • $(E,\mathcal E)$ be a measurable space
  • $(X_n)_{n\in\mathbb N_0}$ and $(Y_n)_{n\in\mathbb N_0}$ be independent $(E,\mathcal E)$-valued time-homogeneous Markov chains on $(\Omega,\mathcal A,\operatorname P)$ with common transition kernel $\kappa$ and $$Z_n:=(X_n,Y_n)\;\;\;\text{for }n\in\mathbb N_0$$
  • $\mathcal F^X$, $\mathcal F^Y$ and $\mathcal F^Z$ denote the filtraiton generated by $X$, $Y$ and $Z$, respectively

It's easy to see that $$\tau:=\inf\left\{n\in\mathbb N_0:X_n=Y_n\right\}$$ is an $\mathcal F^Z$-stopping time and hence $$\tilde Y_n:=1_{\left\{\:n\:<\:\tau\:\right\}}Y_n+1_{\left\{\:n\:\ge\:\tau\:\right\}}X_n\;\;\;\text{for }n\in\mathbb N_0$$ is $\mathcal F^Z$-adapted. Moreover, $\mathcal F^Z=\mathcal F^X\vee\mathcal F^Y$.

How can we show that $\tilde Y$ is a time-homogeneous Markov chain with the same distribution as $Y$?

I guess the basic idea is that $Z$ is clearly a time-homogeneous Markov chain with transition kernel $\pi$ satisfying $$\pi((x,y),B_1\times B_2)=\kappa(x,B_1)\kappa(y,B_2)\;\;\;\text{for all }x,y\in E\text{ and }B_i\in\mathcal E\tag1.$$ Since $\mathbb N_0$ is countable, $Z$ is strongly Markovian at $\tau$ and hence $$1_{\left\{\:\tau\:<\:\infty\:\right\}}\operatorname E\left[f\left(\left(Z_{\tau+n}\right)_{n\in\mathbb N_0}\right)\mid\mathcal F_\tau\right]=1_{\left\{\:\tau\:<\:\infty\:\right\}}(\pi f)(Z_\tau)\;\;\;\text{almost surely}\tag2,$$ where $\pi f:=\int\pi(\;\cdot\;,{\rm d}z)f(z)$, for all bounded and $(\mathcal E\otimes\mathcal E)^{\otimes\mathbb N_0}$-measurable $f:(E\times E)^{\mathbb N_0}\to\mathbb R$. So, if $k\in\mathbb N_0$, $n_0,\ldots,n_k\in\mathbb N_0$ with $0=n_0<\cdots<n_k$ and $B\in\mathcal E^{\otimes k}$, we obtain \begin{equation}\begin{split}1_{\left\{\:\tau\:<\:\infty\:\right\}}\operatorname P\left[\left(\tilde Y_{\tau+n_1},\ldots,\tilde Y_{\tau+n_k}\right)\in B\mid\mathcal F_\tau\right]&=1_{\left\{\:\tau\:<\:\infty\:\right\}}\bigotimes_{i=1}^k\kappa^{n_i-n_{i-1}}(X_\tau,B)\\&=1_{\left\{\:\tau\:<\:\infty\:\right\}}\operatorname P\left[\left(Y_{\tau+n_1},\ldots,Y_{\tau+n_k}\right)\in B\mid\mathcal F_\tau\right]\end{split}\tag3\end{equation} almost surely.

However, it's neither clear to me how we can conclude that $\tilde Y$ is Markovian (with respect to its generated filtration) nor why it has the same distribution as $Y$.

Clearly, the distribution of $Y$ is uniquely determined by the finite-dimensional distributions $\operatorname P\left[\left(Y_{n_1},\ldots,Y_{n_k}\right)\;\cdot\;\right]$ (and the same applies to $\tilde Y$). Moreover, we may write $$\operatorname P\left[\left(\tilde Y_{n_1},\ldots,\tilde Y_{n_k}\right)\;\cdot\;\right]=\operatorname P\left[n<\tau,\left(Y_{n_1},\ldots,Y_{n_k}\right)\;\cdot\;\right]+\operatorname P\left[n\ge\tau,\left(X_{n_1},\ldots,X_{n_k}\right)\;\cdot\;\right]\tag4.$$ Many pieces, I'm not able to combine.

  • $\begingroup$ Do you assume that $X$ and $Y$ are independent? $\endgroup$
    – saz
    Jan 13 '19 at 12:02
  • $\begingroup$ @saz Oops! Somehow forgotten to mention that. Yes, of course. $\endgroup$
    – 0xbadf00d
    Jan 14 '19 at 0:08

Since $X_{\tau} = Y_{\tau}$ on $\{\tau<\infty\}$ it holds that

$$\bar{Y}_n = 1_{\{\tau \geq n\}} Y_n + 1_{\{\tau \leq n-1\}} X_n.$$

Using that $\{\tau \leq n-1\} \in \mathcal{F}_{n-1}^Z$ we find from the pull out property of conditional expectation that

$$\mathbb{E}(f(\bar{Y}_n) \mid \mathcal{F}_{n-1}^Z) = 1_{\{\tau \geq n\}} \mathbb{E}(f(Y_n) \mid \mathcal{F}_{n-1}^Z) + 1_{\{\tau \leq n-1\}} \mathbb{E}(f(X_n) \mid \mathcal{F}_{n-1}^Z)$$

for any bounded measurable function $f$. Since $X$ and $Y$ are, by assumption, independent it follows (see the lemma below) that

$$\mathbb{E}(f(\bar{Y}_n) \mid \mathcal{F}_{n-1}^Z) = 1_{\{\tau \geq n\}} \mathbb{E}(f(Y_n) \mid \mathcal{F}_{n-1}^Y) + 1_{\{\tau \leq n-1\}} \mathbb{E}(f(X_n) \mid \mathcal{F}_{n-1}^X).$$

By assumption, $X$ and $Y$ are both Markov chains with transition kernel $\kappa$, and so

$$\begin{align*} \mathbb{E}(f(\bar{Y}_n) \mid \mathcal{F}_{n-1}^Z) &= 1_{\{\tau \geq n\}} \int f(y) \, \kappa(Y_{n-1},dy) + 1_{\{\tau \leq n-1\}} \int f(y) \, \kappa(X_{n-1},dy) \\ &= \int f(y) \, \kappa(\bar{Y}_{n-1},dy). \end{align*}$$

Since $n \in \mathbb{N}$ is arbitrary, this shows that $(\bar{Y}_n)_{n \in \mathbb{N}}$ is a Markov chain with transition kernel $\kappa$.

Lemma Let $Z \in L^1(\mathbb{P})$ be a random variable which is measurable with respect to a $\sigma$-algebra $\mathcal{A}$. If $\mathcal{G},\mathcal{H}$ are further $\sigma$-algebras such that $\mathcal{H}$ is independent from $\sigma(\sigma(Z),\mathcal{G})$, then $$\mathbb{E}(Z \mid \sigma(\mathcal{G},\mathcal{H})) = \mathbb{E}(Z \mid \mathcal{G}).$$

  • $\begingroup$ Your lemma is just that if $\mathcal F$ is independent of $\mathcal G\vee\mathcal H$, then $\mathcal F$ is conditionally independet of $\mathcal H$ given $G$, right? That is clear since $\left\{G∩H:G∈\mathcal G\text{ and }H∈\mathcal H\right\}$ is a $\cap$-stable generator of $\mathcal G\vee\mathcal H$ and $\text P[F∩G∩H]=\text P[F]\text P[G∩H]=\text E[1_F]\text E[1_G\text P[H\mid\mathcal G]]=\text E[1_{F∩G}\text P[H\mid\mathcal G]]$ for all $F∈\mathcal F$, $G∈\mathcal G$ and $H∈\mathcal H$ (since $F$ is independent of $G∩H$ and $F$ is independent of $\mathcal G$). $\endgroup$
    – 0xbadf00d
    Jan 14 '19 at 22:32
  • $\begingroup$ @0xbadf00d Note that $\mathcal{H}$ is assumed to be independent, not $\mathcal{F}$. It might well be that it is possible to formulate the lemma in terms of conditional independence but I don't see the point in it... for the proof I need the lemma the way I formulated it. $\endgroup$
    – saz
    Jan 15 '19 at 7:38

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.