2
$\begingroup$

Define a coupling $\pi\in \Pi(\mu,\nu)$ on the product space $(X\times X,\mathcal{F}\times\mathcal{F})$. let $\pi_x$ be the disintegration of $\pi$ with respect to the $\mu$, i.e. there exists a Borel measurable function $x\mapsto\pi_x$ such that $$\pi(dx,dy)=\mu(dx)K(x,dy)$$ where $K(\cdot,\cdot)$ is a probability transition kernel.

If $\pi_x=\delta_{f(x)}$ ($\delta$ is the Dirac measure), can we say that there exists a measurable function $f$ such that $$ \nu=f_{\#}\mu $$ where $f_{\#}\mu$ is the pushforward of measure $\mu$, i.e. $\mu(f^{-1}(A))=\nu(A)$ for any Borel measurable sets $A\subset X$.

$\endgroup$

1 Answer 1

2
$\begingroup$

Yes: if $\pi_x=\delta_{f(x)}$, then $$\nu(A)=\pi(X\times A)=\int_X\mu(dx)\pi_x(A) \\ =\int_X\mu(dx)\,1(f(x)\in A)=\int_X\mu(dx)\,1(x\in f^{-1}(A)) =\mu(f^{-1}(A))$$ for all Borel subsets $A$ of $X$, as desired.

$\endgroup$
2
  • $\begingroup$ Why is $f$ measurable? $\endgroup$ Jul 17, 2022 at 19:37
  • $\begingroup$ @DieterKadelka : This follows because $(\pi_x)$ is a disintegration of $\pi$, which implies that all the equalities in my answer must make sense. If $f$ were not measurable, then the latter integral there, as well as $\mu(f^{-1}(A))$, would not make sense for some Borel $A$. $\endgroup$ Jul 17, 2022 at 19:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.