A Treatise on Measurability

Throughout this writeup, let \Omega, \Psi be sets and X : \Omega \to \Psi be a function. Refer to a previous post for prior measure-theoretic notions that will be used in this set of exercises.

Problem 1. For any \sigma-algebra \mathcal G on \Psi, define \mathcal G_{X^{-1}} by

\mathcal G_{X^{-1}} := \{X^{-1}(M) : M \in \mathcal G\}.

Prove that \mathcal G_{X^{-1}} is a \sigma-algebra on \Omega, known as the pull-back \sigma-algebra of X.

(Click for Solution)

Solution. We have \emptyset = X^{-1}(\emptyset) \in \mathcal G_{X^{-1}} and \Omega = X^{-1}(\Psi) \in \mathcal G_{X^{-1}}.

The complementation and union properties follows from set algebra via

\displaystyle \begin{aligned} \Omega \backslash X^{-1}(M) &= X^{-1} (\Omega \backslash M)  \in \mathcal G_{X^{-1}}, \\ \bigcup_{i=1}^\infty X^{-1}(V_i) &= X^{-1} \left( \bigcup_{i=1}^\infty V_i \right) \in \mathcal G_{X^{-1}}. \end{aligned}

Problem 2. For any \sigma-algebra \mathcal F on \Omega, define \mathcal F_X by

\mathcal F_X := \{M \in \mathcal P(\Psi) : X^{-1}(M) \in \mathcal F\}.

Prove that \mathcal F_X is a \sigma-algebra on \Psi, known as the push-forward \sigma-algebra of X.

(Click for Solution)

Solution. Follow the proof in Problem 1.

Let \mathcal F be a \sigma-algebra on \Omega and \mathcal G be a \sigma-algebra on \Psi.

Definition 1. A function W : \Omega \to \Psi is said to be \mathcal F/\mathcal Gmeasurable if for any M \in \mathcal G, W^{-1}(V) \in \mathcal F.

Problem 3. Prove that X is \mathcal F/\mathcal G-measurable if and only if \mathcal G_{X^{-1}} \subseteq \mathcal F. Since this holds for any \sigma-algebra \mathcal F, we say that \mathcal G_{X^{-1}} is the coarsest \sigma-algebra such that X is measurable.

(Click for Solution)

Solution. We first prove (\Rightarrow). Suppose X is \mathcal F/\mathcal G-measurable. Fix L \in \mathcal G_{X^{-1}}. Then L = X^{-1}(M) for some M \in \mathcal G, which implies L = X^{-1}(M) \in \mathcal F, as required.

Next, we prove (\Leftarrow). Suppose \mathcal G_{X^{-1}} \subseteq \mathcal F. Fix M \in \mathcal G. By definition, X^{-1}(G) \in \mathcal G_{X^{-1}} \subseteq \mathcal F. Therefore, X is \mathcal F/\mathcal G-measurable.

Problem 4. Prove that X is \mathcal F/\mathcal G-measurable if and only if \mathcal G \subseteq \mathcal F_X. Since this holds for any \sigma-algebra \mathcal G, we say that \mathcal F_X is the finest \sigma-algebra such that f is measurable.

(Click for Solution)

Solution. We first prove (\Rightarrow). Suppose X is \mathcal F/\mathcal G-measurable. Fix M \in \mathcal G. By measurability, X^{-1}(M) \in \mathcal F, so that M \in \mathcal F_X.

Next, we prove (\Leftarrow). Suppose \mathcal G \subseteq \mathcal F_X. Fix M \in \mathcal G. By assumption, M \in \mathcal G \subseteq \mathcal F_X, so that X^{-1}(M) \in \mathcal F. Therefore, X is \mathcal F/\mathcal G-measurable.

Problem 5. For any \Lambda \subseteq \Omega, prove that \mathcal F_{\iota^{-1}} = \{\Lambda \cap L : L \in \mathcal F\}, where \iota denotes the inclusion map.

(Click for Solution)

Solution. Each element in \mathcal F_{\iota^{-1}} is of the form \iota^{-1}(L) for some L \in \mathcal F. Since \iota^{-1}(L) \subseteq \Lambda and \iota^{-1}(L) \subseteq L, we have \iota^{-1}(L) \subseteq \Lambda \cap L. The result follows by noticing \iota(\Lambda \cap L) \subseteq L \Rightarrow \Lambda \cap L \subseteq \iota^{-1}(L).

Problem 6. Let (\Phi, \mathcal H) be a measurable space. Suppose Y : \Psi \to \Phi is \mathcal G/\mathcal H-measurable. Prove that Y\circ X : \Omega \to \Phi is \mathcal F/\mathcal H-measurable.

(Click for Solution)

Solution. Fix N \in \mathcal H. Since Y is \mathcal G/\mathcal H-measurable, Y^{-1}(W) \in \mathcal G. Since X is \mathcal F/\mathcal G-measurable, (Y \circ X)^{-1}(N) = X^{-1}(Y^{-1}(N)) \in \mathcal F. Therefore, g \circ f îs \mathcal F/\mathcal H-measurable.

—Joel Kindiak, 9 Feb 25, 0107H

,

Published by


Leave a comment