2
$\begingroup$

Suppose that $X_{1}, X_{2}, \ldots, X_{n}$ are random variables with $$X_{i}(\omega) : (\Omega, \mathcal{F})\to (\mathbb{R},\mathcal{R})$$ where $\Omega$ is the sample space, $\mathcal{F}$ is the $\sigma$-field, and $\mathcal{R}$ is the Borel $\sigma$-field.

Now suppose that $f(X_{1}, X_{2}, \ldots, X_{n})$ is a measurable function, $f:(\mathbb{R}^{n}, \mathcal{R}^{n})\to (\mathbb{R},\mathcal{R})$, and consider the $\sigma$-fields generated by the random variables $X_{1}, X_{2}, \ldots, X_{n}$, which I will denote $\sigma(X_{1}), \sigma(X_{2}), \ldots, \sigma(X_{n})$. Finally, let $Y = f(X_{1}(\omega), X_{2}(\omega), \ldots, X_{n}(\omega))$ be a random variable with $Y: (\Omega, \mathcal{F}) \to (\mathbb{R},\mathcal{R})$.

I want to show that $$\sigma(Y) \subset \sigma\left(\bigcup_{i=1}^{n} \sigma(X_{i})\right)$$

I am having trouble with the exact formulation of the proof.

Can anyone provide a rigorous proof of this?

Motivation:

In Durrett's Probability: Theory and Examples (fourth edition), the proof of theorem 2.1.6. basically says that if $\mathcal{G} = \sigma\left(\bigcup_{j} \sigma(X_{j})\right)$, then $f(X_{1}, X_{2}, \ldots, X_{n}) \in \mathcal{G}$ (I believe Durrett meant to put $f^{-1}(X_{1}, X_{2}, \ldots, X_{n}) \in \mathcal{G}$). I want to make sense of why this is true.

$\endgroup$
  • $\begingroup$ In the current form, the question doesn't make sense. $\sigma(f)$ is a $\sigma$-algebra in $\mathbb{R}^n$ and $\sigma(X_i)$ is a $\sigma$-algebra in $\Omega$. $\endgroup$ – Dominik Dec 5 '16 at 15:55
  • $\begingroup$ @Dominik Perhaps what I wrote did not reflect what I had in mind. I have revised the question $\endgroup$ – möbius Dec 5 '16 at 16:00
1
$\begingroup$

edit: This first answer was for the first version of the quesion, see below for the second version.

Your claim is false in general:

Take $X_i \equiv 0$ then $\sigma(X_i)$ is trivial for each $i$ hence $$\sigma\left(\bigcup_{i=1}^{n} \sigma(X_{i})\right) = \{\emptyset,\Omega\}$$.

Do you really want to claim that $\sigma(f) = \{\emptyset,\Omega\}$ for each measurable function $f$? :-)

Let $X = (X_1,\ldots,X_n)$ then $Y = f\circ X$ so $Y^{-1} = X^{-1} \circ f^{-1}$

Hence $$\sigma(Y) = \sigma(\{ (X^{-1} \circ f^{-1})(A) | A \in \mathcal{R}\}) = \sigma(\{ (X^{-1}\left(f^{-1}(A)\right) | A \in \mathcal{R}\})$$

And $$\sigma\left(\bigcup_{i=1}^{n} \sigma(X_{i})\right) = \sigma(X_i^{-1}(A) | A \in \mathcal{R}, i\in\{1,\ldots,n\})$$

Because $f$ is measurable it holds $f^{-1}(A) \in \mathcal{R}$ so $$\sigma(Y) \subseteq \sigma\left(\bigcup_{i=1}^{n} \sigma(X_{i})\right)$$

$\endgroup$
  • $\begingroup$ Question about the last step: I agree $f^{-1}(A) \in \mathcal{R}$. For simplicity let $B=f^{-1}(A)$. Then basically we want to show that $\{X^{-1}(B) | B\in\mathcal{R}\} \subset \{X_{i}^{-1}(B) | B\in\mathcal{R}, i\in \{1,\ldots,n\}\}$. Can you elaborate on how to do this final step? $\endgroup$ – möbius Dec 5 '16 at 16:18
  • $\begingroup$ T.b.h you have to do some more paperwork in there due to the fact that $X^{-1}(B) \subseteq \Bbb R^n$ but $X_i^{-1}(B) \subseteq \Bbb R$ or more detailed $X^{-1}(B) \in \mathcal{R}^n$ but $X_i^{-1}(B) \in \mathcal{R}$. So to be correct you should state $\sigma(Y) \subseteq \sigma\left((\sigma(X_1),\sigma(X_2),\ldots,\sigma(X_n)\right)$ $\endgroup$ – Gono Dec 5 '16 at 16:29

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.