Let $X$ and $Y$ be identically distributed bounded random variables defined on a probability space $(\Omega,\mathcal{F}, \mathbb{P})$. I want to know if there always exists an invertible measure-preserving transformation $T: \Omega \to \Omega$ such that $ X \circ T= Y$ holds almost surely.

When $X$ and $Y$ are characteristic functions, this is essentially proven in page 74 of Halmos's Lectures on Ergodic theory. Before attempting to prove the general statement myself, I want to ask if this is possibly a standard result. I would very much appreciate any suggestions or references.


No. Take for example the case where $\Omega = [0,1]$ with Lebesgue measure, $X$ is the identity, and $Y = 2X\mod 1$. Then you ask for $T = X\circ T = Y$ almost surely, but changing $Y$ on a set of measure zero cannot make it invertible.

  • $\begingroup$ Thank you. What if if I do not insist on $T$ being invertible? $\endgroup$ – Keivan Karai Jul 19 '13 at 23:10
  • $\begingroup$ @KeivanKarai: I believe in that case you can get a counterexample by letting $X(\omega) = 2\omega \mod 1$ and $Y(\omega) = 3\omega\mod 1$, though I haven't checked the details. $\endgroup$ – Noah Stein Jul 22 '13 at 1:21

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.