1
$\begingroup$

On Pg. 5 of Probability and Measure Theory the authors argue as follows:

If $\mathscr{C}$ is a class of subsets of $\Omega$ and $A \subset \Omega$, we denote by $\mathscr{C} \cap A$ the class $\{ B \cap A : B \in \mathscr{C}\}$. If the minimal $\sigma$-field over $\mathscr{C}$ is $\sigma(\mathscr{C}) = \mathscr{F}$ then the minimal $\sigma$-field of subsets of $A$, $\sigma_A(\mathscr{C} \cap A) = \mathscr{F} \cap A$.

They build the proof by arguing that $\mathscr{F} \cap A$ is a $\sigma$-field of subsets of $A$.

I don't see how that is obvious. First, $\mathscr{F}$ is a $\sigma$-field of $\mathscr{C}$ which has no relation to $A$. Second, a class of sets formed by intersection with $A$ is not guaranteed to contain $A$. So how is it that $\mathscr{F} \cap A$ is a $\sigma$-field of subsets of $A$?

Or did I misunderstand something here?

$\endgroup$
4
$\begingroup$

When the authors say "the minimal $\sigma$-field over $\mathscr{C}$", they mean the smallest $\sigma$-field of $\Omega$ containing $\mathscr{C}$ [they explain their terminology in the middle of page 4].

$\mathscr{F}\cap A$ contains the emptyset since $\mathscr{F}$ does. It contains $A$ since $\mathscr{F}$ contains $\Omega$. It is closed under complements since $A\setminus (F\cap A) = (\Omega\setminus F) \cap A$ for each $F \in \mathscr{F}$. It is closed under countable unions since $\cup_n (F_n\cap A) = (\cup_n F_n) \cap A$ for each collection $\{F_n\} \in \mathscr{F}$. It is closed under countable intersections since it is closed under countable unions and taking complements.

$\endgroup$
  • $\begingroup$ That’s probably where I am confused. How come $\mathscr{F}$ contain $\Omega$? Is $\mathscr{C}$ “the” class of subsets of $\Omega$ or “a” class of subsets of $\Omega$? $\endgroup$ – Apoorv Khurasia Aug 5 '18 at 9:07
  • $\begingroup$ @ApoorvKhurasia $\mathscr{C}$ is a class of subsets. $\mathscr{F}$ is the smallest sigma-field containing $\mathscr{C}$. In particular, $\mathscr{F}$ is a sigma-field, so it contains $\Omega$. My advice would be to ignore $\mathscr{C}$ and note that my answers shows the general claim that if $\mathscr{F}$ is a sigma-field, then $\mathscr{F}\cap A$ is a sigma-field for any $A \subseteq \Omega$. $\endgroup$ – mathworker21 Aug 5 '18 at 9:25
  • $\begingroup$ how do we know $\Omega \in \mathscr{F}$? We only know that $\mathscr{F}$ is a $\sigma$-field over $\mathscr{C}$ not $\Omega$ with $\mathscr{C}$ being an arbitratry class of subsets of $\Omega$. Or perhaps I am missing something very fundamental. $\endgroup$ – Apoorv Khurasia Aug 6 '18 at 8:18
  • $\begingroup$ I don't know what "$\mathscr{F}$ is a $\sigma$-field over $\mathscr{C}$" means. What I was assuming is that $\mathscr{F}$ is the smallest $\sigma$-field (over $\Omega$) containing $\mathscr{C}$. $\endgroup$ – mathworker21 Aug 6 '18 at 8:24
  • 1
    $\begingroup$ @ApoorvKhurasia Look at the middle of page 4, where the author defines "minimal sigma-field over". That explains his (weird) terminology. $\endgroup$ – mathworker21 Aug 6 '18 at 10:18
2
$\begingroup$

In general if $f:X\to Y$ is a function and $\mathcal A$ denotes a $\sigma$-algebra on $Y$ then it can be shown that $f^{-1}(\mathcal A):=\{f^{-1}(A)\mid A\in\mathcal A\}$ is a $\sigma$-algebra on $X$.

Preimages are very coöperative by proving that (give it a try).

Your case is a special case where $\mathscr F=i^{-1}(\mathscr C)$ where $i:A\to\Omega$ denotes the inclusion.

It is a good thing to be familiar the more general rule.


There is also a rule that is even more general. I will not handle that here, but in case you are interested here you can find a proof of that.

$\endgroup$
  • $\begingroup$ I finally managed to prove this and I found this theorem to be handy. Thanks. $\endgroup$ – Apoorv Khurasia Aug 12 '18 at 6:29
  • $\begingroup$ Nice to hear that. You are welcome. $\endgroup$ – drhab Aug 12 '18 at 7:25

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.