Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

If a random variable $X$ has a density $f$, then the expected value can be simplified: $$\mathbb{E}[X]=a+∫_{a}^{b}(1-F(x))dx,$$

where $F$ is the cumulative distribution function, $F(x)=\Pr(X≤x)$.

My question is: Does this simplification should work for all probability distributions, even for those that are not absolutely continuous with respect to Lebesgue-measure on $[a,b]$?

If $X$ is any real-valued random variable with support $[a,b]$ and $F(x)=\Pr(x≤X)$, is it always true that $$ \mathbb{E}[X]=a+∫_{a}^{b}(1-F(x))dx$$

The answer to this question would be simple if one could generally extend integration-by-parts to Lebesgue-Stieltjes integration. However, this is not possible; see Wikipedia.

share|cite|improve this question
It is better to write capital $X$ for the random variable and lower-case $x$ for the argument to the c.d.f. or the p.d.f., instead of using the same thing for both usages. Among other advantages, you avoid the awkwardness of such expressions as "$\mathbb{P}(x \le x\prime)$", instead writing $\Pr(X\le x)$. (And notice that the code for $\Pr$ is \Pr.) – Michael Hardy Jun 12 '12 at 0:07
It does work for all Borel-measurable distributions that are supported on some subset of $[a,\infty)$, where $a$ is finite. I don't think you can do it if the support has no finite lower bound. – Michael Hardy Jun 12 '12 at 0:10
up vote 1 down vote accepted

Let's use $X$ for the random variable, keeping $x$ for the variable of integration. In general, we have a probability measure $\mu$ on $I = [a,b]$ and

$$\eqalign{E[X] &= \int_I x\ d\mu(x) = a + \int_I (x-a) d\mu(x)\cr &= a + \int_I \int_a^x 1 \ dt \ d\mu(x) = a + \int_a^b \int_{[t,b]} 1 \ d\mu(x)\ dt \cr &= a + \int_a^b (1 - F(t)) \ dt\cr}$$

(note that $\int_{[t,b]} 1 \ d\mu(x) = 1 - F(t-)$, but that is $1 - F(t)$ (Lebesgue) almost everywhere)

share|cite|improve this answer

"The answer to this question would be simple if one could generally extend integration-by-parts to Lebesgue-Stieltjes integration."

This comment moves me to add another answer: Riemann--Stieltjes integration is enough for this. Look it up and/or make it an exercise to prove it. I.e. suppose you have a probability distribution on the set of Borel-measurable subsets of $\mathbb{R}$. Let $X$ be a random variable so distributed and let $F$ be the c.d.f., so $F(x)=\Pr(X\le x)$. Then $$ \mathbb{E}(X) = \int_{-\infty}^\infty x \, dF(x), $$ where that is a Riemann--Stieltjes integral. If the support of the distribution has a finite a lower bound, then applying integration by parts to the Riemann--Stieltjes integral does it.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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