Let $X$ be a real non-negative random variable on the probability space $(\Omega,\mathcal{F}, \mathbb{P})$. Given that $$ E[X]=\int_\Omega \int_0^\infty \chi_{t<X}\,dt\,d\mathbb{P}=\int_0^\infty \mathbb{P}[ X\geq t]\,dt, $$ show that, for all $\epsilon>0$ $$ E[X]\leq \sum_{n=0}^\infty \epsilon\mathbb{P}[X\geq n\epsilon]\leq E[X]+\epsilon. $$ Tried to use some Fubini combined with rewriting stuff as countable sums (like $\mathbb{P}[X\geq t]=\mathbb{P}\left[\bigcup_{n=1}^\infty \{X\geq nt\}\right]$) but I am a bit lost. Some intuition is also highly appreciated (I think I am beaten by the misunderstanding of notation).


Notice that the summation in the middle is the Riemann sum of the decreasing function $f(t) = P(X>t).$ The monotonicity makes it easy to estimate the difference between the sum and the integral. Now, the question is: is the function Riemann integrable, and if not, does it matter?

  • $\begingroup$ I would say that we do not know whether the probability measure of $X$ is Riemann integrable (is this what you are talking about?). However calling it a Riemann sum isn't a stronger assumption given that the trick can be done within the Lebesgue integration (doesn't it?). $\endgroup$ – Ton Jan 27 '14 at 23:15
  • 1
    $\begingroup$ @Ton It is actually NOT the probability measure, but the CDF. Part of the point is that this is a monotonic function and therefore IS Riemann-integrable, so... $\endgroup$ – Igor Rivin Jan 28 '14 at 1:06

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.