Suppose $(\mathfrak{B}_n)_n$ are independent with respect to a finite measure $\mu$. Show that for any $N$, the $\sigma$-algebras $\mathfrak{B}_1,\mathfrak{B}_2,...,\mathfrak{B}_N,\sigma(\bigcup_{n=N+1}^\infty\mathfrak{B}_n)$ are also independent.

It is enough to show that $\mu(A\cap B)=\mu(A) \mu(B)$ where $A\in\mathfrak{B}_i$ for some $i\in\{1,...,N \}$ and $B\in \sigma(\bigcup_{n=N+1}^\infty\mathfrak{B}_n)$. I know also that $\sigma(\bigcup_{n=N+1}^\infty\mathfrak{B}_n)=\sigma(\epsilon)$ where $\epsilon=\bigcup_{n=1}^\infty\epsilon_n$ and $\epsilon_n=\{\bigcap_{k=N+1}^n A_k: A_{N+1}\in\mathfrak{B}_{N+1},A_{N+2}\in\mathfrak{B}_{N+2},...,A_{n}\in\mathfrak{B}_{n}\}$ But I couldn't continue because I don't know how to write an element $B\in \sigma(\epsilon)$. Thanks for your helps!


Let $\mathscr{R}$ be the class of sets $B$ such that, for all $A\in\mathfrak{B}_1\cup\cdots\cup\mathfrak{B}_N$, we have $\mu(A\cap B)=\mu(A)\mu(B)$.

Clearly, $\bigcup_{k=N+1}^\infty\mathfrak{B}_k\subseteq\mathscr{R}$. I believe you can show by yourself that $\mathscr{R}$ is a $\sigma$-algebra (you're going to use finiteness of $\mu$ when dealing with complements), thus $\sigma(\bigcup_{k=N+1}\mathfrak{B}_k)\subseteq\mathscr{R}$. This means that for all $B\in\sigma(\bigcup_{k=N+1}^\infty\mathfrak{B}_k)$ and all $A\in\mathfrak{B}_j$ for $j\leq N$, $\mu(A\cap B)=\mu(A)\mu(B)$.

Remark Note that we have to assume that $\mu$ is in fact a probability. If we have two $\sigma$-algebras $\mathfrak{B}_1$ and $\mathfrak{B}_2$ on a space $X$ which are independent with respect to a finite measure $\mu$, we have $X\in\mathfrak{B}_1$ and $X\in\mathfrak{B}_2$, so $\mu(X)=\mu(X\cap X)=\mu(X)\mu(X)$, thus either $\mu(X)=0$ or $1$.


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.