I have a question about the definition of a Markov process and a Markov family by Karatzas/Shreve in the book "Brownian Motion and Stochastic Calculus" (cf. p. 74). Let me first recall their definitions:

Markov process: Let $d$ be a positive integer and $\mu$ a probability measure on $(\mathbb{R}^d, \mathcal{B}(\mathbb{R}^d))$. An adapted, $d$-dimensional process $X=\{X_t,\mathcal{F}_t;t\geq 0 \}$ on some probability space $(\Omega, \mathcal{F},P^{\mu})$ is said to be a Markov process with initial distribution $\mu$, if

(i) $P^{\mu}(X_0 \in \Gamma ) = \mu(\Gamma), \forall \Gamma \in \mathcal{B}(\mathbb{R}^d)$.

(ii) for $s,t \geq 0$ and $\Gamma \in \mathcal{B}(\mathbb{R}^d)$: $P^{\mu} (X_{t+s} \in \Gamma | \mathcal{F}_s) = P^{\mu} (X_{t+s} \in \Gamma | X_s)$

Markov Family: Let $d$ be a positive integer. A $d$-dimensional Markov Family is an adapted process $X=\{X_t,\mathcal{F}_t;t\geq 0 \}$ on some $(\Omega, \mathcal{F})$, together with a family of probability measures $(P^x)_{x\in \mathbb{R}^d}$ on $(\Omega,\mathcal{F})$, such that

(a) for each $F\in \mathcal{F}$, the mapping $x\mapsto P^x(F)$ is universally measurable;

(b) $P^x(X_0=x)=1, \forall x \in \mathbb{R}^d$;

(c) for $x\in \mathbb{R}^d$, $s,t \geq 0$ and $\Gamma \in \mathcal{B}(\mathbb{R}^d)$: $P^x(X_{t+s} \in \Gamma | \mathcal{F}_s ) = P^x(X_{t+s} \in \Gamma | X_s ), P^x-a.s.$;

(d) for $x\in \mathbb{R}^d$, $s,t \geq 0$ and $\Gamma \in \mathcal{B}(\mathbb{R}^d)$: $P^x(X_{t+s} \in \Gamma | X_s=y ) = P^y(X_{t} \in \Gamma), P^xX_s^{-1}-a.e. y$;

Now my question: To me it looks like the definition of a Markov Family serves to have a Markov process start at different starting points, and for each different starting point $x$, we have one measure $P^x$. Why do we introduce this new definition instead of considering a family of processes $(X^x_t)_{x\in \mathbb{R}^d, t\geq 0}$ with $X^x_t = X_t+x$, and $(X_t)_{t\geq 0}$ a Markov process starting in zero with respect to one measure $P^0$. In this case we would have a family of processes but only one measure, whereas in the aforementioned definition we have one process but a family of measures. What is the advantage of the aforementioned definition?

Thank you in advance, Luke


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.