# What does it mean for a stochastic process to be independent of a sigma algebra?

Does anybody know what it means for a stochastic process $X = (X_t)_{t \geq 0}$ on a filtered probability space $(\Omega, \mathcal{F}, \mathbb{F}, P)$ to be independent of a sigma-algebra $\mathcal{G} \subset \mathcal{F}$ ?

Is this the same as saying that $\forall 0 \leq t_1 < ... < t_n$ the vector $(X_{t_1}, ... , X_{t_n})$ is independent of $\mathcal{G}$ ?

Thanks for the clarification!

Regards, Si

-

I think the natural definition would be that the $\sigma$-algebra $\sigma(X_t : t\ge 0)$ generated by the process should be independent of $\mathcal{G}$. That is, for every $A \in \sigma(X_t : t \ge 0)$ and every $B \in \mathcal{G}$, $P(A \cap B) = P(A) P(B)$.

However, this is equivalent to your statement. One can use Dynkin's $\pi$-$\lambda$ lemma to prove it.

-

We can more generally define the independence between two processes $X=(X_t)_t$ and $Y=(Y_t)_t$ by the following way: if $\mathcal F_X$ and $\mathcal F_Y$ are the smallest $\sigma$-algebra making respectively $X$ and $Y$ measurable, we say that $X$ and $Y$ are independent if and only if so are $\mathcal F_X$ and $\mathcal F_Y$.

We can show the equivalence between

• $X$ and $Y$ are independent;
• The finite-dimensional distributions of $X$ and $Y$ are independent.

If we assume that the finite-dimensional distributions of $X$ and $Y$ are independent, then consider $\pi_1$ and $\pi_2$, the sets which consists of finite intersection of sets of the form $X_t^{-1}(B)$ (respectively $Y_t^{-1}$), where $B$ is a Borel set. These one are $\pi$-systems (i.e. stable by finite intersection), and a monotone class result ensure us that the monotone class generated by $\pi$ and $\pi_2$ are independent.

-
Many thanks to both of you! –  Mad Si May 1 '12 at 15:12
You are welcome! –  Davide Giraudo May 1 '12 at 16:20