Problem. Let $\left(\Omega, \mathcal A, \Bbb{P} \right)$ be a probability space and $X,Y$ be real-valued random variables with finite expected value. Then $$E(Y)-E(X)= \int_\Bbb{R}\Bbb{P}(X<t\leq Y)-\Bbb{P}(Y<t\leq X)\, dt.$$

My attempt. This isn't much of an attempt, but I've written $$\int_\Bbb{R}\Bbb{P}(X<t\leq Y)-\Bbb{P}(Y<t\leq X)\, dt = \int_\Bbb{R} E\left(\mathbf 1\{X<t\leq Y\}-\mathbf 1\{Y<t\leq X\} \right) dt$$ Can we interchange integration and $E$? How could we proceed?


Yes, we interchange integration with expectation using Fubini-Tonelli. (This uses the finite expected value condition.) $$ \int_{\mathbb R}E({\bf 1}\{X<t\leq Y\}-{\bf 1}\{Y<t\leq X\})=E\left(\int_{\mathbb R}{\bf 1}\{X<t\leq Y\}-{\bf 1}\{Y<t\leq X\}\right)=E(Y-X). $$ The last step follows from the the deterministic identity $$ Y-X=\int_{\mathbb R}{\bf 1}\{X<t\leq Y\}-{\bf 1}\{Y<t\leq X\} $$ for any real numbers $X,Y\in \mathbb R$ (just consider each of the two cases $X\leq Y$ and $X\geq Y$ separately).

  • $\begingroup$ why Tonelli? is everything is positive here? $\endgroup$ – Math-fun Sep 9 '15 at 8:26
  • $\begingroup$ Well, the standard way is to split $f$ into positive and negative parts and to apply Tonelli to each separately. $\endgroup$ – pre-kidney Sep 9 '15 at 8:28

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.