Sectioning sets in product measures

55 Views Asked by At

I'm reading through Capinski and Kopp's Measure, Integral and Probability and stumbled across Theorem 6.4 about sectioning sets in product measures:

If $A$ is in the product $\sigma$-field $\mathcal{F} = \mathcal{F}_1 \times \mathcal{F}_2$, then for each $\omega_2$, $A_{\omega_2} \triangleq \left\{ \omega_1 \in \Omega_1 \ \middle|\ (\omega_1, \omega_2) \in A\right\} \in \mathcal{F}_1$ (similarly for $\omega_1$, $A_{\omega_1} \in \mathcal{F}_2$).

Intuitively, this makes sense, however the proof is a bit puzzling: the authors define the set

$$ \mathcal{G} = \left\{ A \in \mathcal{F} \ \middle|\ A_{\omega_2} \in \mathcal{F}_1, \forall \omega_2 \right\} $$ and proceed to show that $\mathcal{G} = \mathcal{F}$ by showing that $\mathcal{G}$ is a $\sigma$-field containing the "rectangles" of $\mathcal{F}_1 \times \mathcal{F}_2$. However, the first step reads:

Take $A = A_1 \times A_2, A_1 \in \mathcal{F}_1$. In that case: $$ A_{\omega_2} = \begin{cases} A_1, & \text{ if } \omega_2 \in A_2 \\ \emptyset, & \text{ otherwise} \end{cases} $$ is in $\mathcal{F}_1$ (since $A_1, \emptyset \in \mathcal{F}_1$) so the rectangles are in $\mathcal{G}$ (i.e. $A \in \mathcal{G}$).

Question: is that step not sufficient for concluding that $A_{\omega_2} \in \mathcal{F}_1$? Why do we need to go through the process of proving that $\mathcal{G}$ is a $\sigma$-field and then concluding $\mathcal{G} = \mathcal{F}$ to say that?

1

There are 1 best solutions below

1
On BEST ANSWER

The step you highlighted only proves that $A_{\omega_2}$ is in $\mathcal F_1$ in the special case where $A$ is a rectangle (i.e. when $A$ can be written as a product, $A_1 \times A_2$, where $A_1 \in \mathcal F_1$ and $A_2 \in \mathcal F_2$). But many measurable sets $A$ in $\mathcal F$ are not rectangles!

[For example, consider endowing $\mathbb R^2 = \mathbb R \times \mathbb R$ with the product measure induced by the Lebesgue measure on each of the two $\mathbb R$'s. Then then disk $\{(x,y) \in \mathbb R^2 : \sqrt{x^2 + y^2 } < 1 \}$ is measurable w.r.t. the product measure, but it is not a rectangle.]

So to go from the easy special case of rectangles to the general case of all measurable sets, the authors structure their argument as follows:

(i) $A_{\omega_2} \in \mathcal F_1$ in the special case where $A$ is a "rectangle" (i.e. a set of the form $A_1 \times A_2$ where $A_1 \in \mathcal F_1$ and $A_2 \in \mathcal F_2$).

(ii) The collection of sets $A\in\mathcal F$ such that $A_{\omega_2} \in \mathcal F_1$ is a sigma-algebra, i.e. it contains the empty set, and it is closed under taking complements and countable unions.

(iii) Since $\mathcal F$ is, by definition, the smallest sigma-algebra containing all rectangles, it must be the case that the collection of sets $A \in\mathcal F$ such that $A_{\omega_2} \in \mathcal F_1$ is the whole of $\mathcal F$.