Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

This is an exercise mentioned in this question:

Let $Y$ be an integrable random variable on the space $(\Omega,{\mathcal A},{\bf P})$ and $\mathcal{G}$ be a sub $\sigma$-algebra of $\mathcal{A}$. Show that $|Y|\leq c$ implies $|E[Y\mid{\mathcal G}]|\leq c$.

It has been shown that using $E[|y|\mid{\mathcal G}]$ to estimate $|E[Y \mid{\mathcal G}]|$ is not a good idea due to did's answer to the previous question.

One more step I can go so far is $E[Y]=E[E[Y\mid{\mathcal G}]]$. I don't think this would work since expectation only gives the average.

So how can I get $|E[Y\mid{\mathcal G}]|\leq c$ using $|Y|\leq c$?

share|cite|improve this question
up vote 2 down vote accepted

First $$ \left|E[Y\mid\mathcal G]\right|\leq E[|Y|\mid \mathcal G] $$ by Jensen's inequality. Now, let $Z=E[|Y|\mid \mathcal G]$, then for every $G\in\mathcal G$ we have $$ \int_G Z=\int_G|Y|\,\mathrm dP\leq \int_G c\;\mathrm dP, $$ and since both $Z$ and $c$ are non-negative and $\mathcal{G}$-measurable, we have that $Z\leq c$ almost surely.

To see that the last statement is true, we consider the non-negative random variable $(Z-c)1_{\{Z>c\}}$ which has mean $0$ as $$ 0\leq E\left[(Z-c)1_{\{Z>c\}}\right]= \int_{\{Z>c\}} Z\,\mathrm dP-\int_{\{Z>c\}} c\,\mathrm dP\leq 0 $$ since ${\{Z>c\}}\in\mathcal{G}$. This implies $(Z-c)1_{\{Z>c\}}=0$ a.s. and hence $Z=c$ a.s.

share|cite|improve this answer
thank you for your answer. how do you come up with the last sentence? – Jack Nov 29 '12 at 21:27
@Jack: See the edit. – Stefan Hansen Nov 30 '12 at 6:24

Let $F_n:=\{E[Y\mid\cal G]\geqslant c+n^{-1}\}\in\cal G$. Then $$c\mathbf P(F_n)\geqslant (c+n^{-1})\mathbf P(F_n),$$ so $\mathbf P(F_n)=0$ for all $n$, hence $E[Y\mid\cal G]\leqslant c$ a.s. Doing the same for $-Y$, we get the wanted result.

share|cite|improve this answer

If $P(Z \leq X)=1$ then $E[Z \vert \mathcal{G}]\leq E[X\vert \mathcal{G}])$ a.s.

Use $Z=|Y|$ and that $c$ is a $\mathcal{G}$-measurable stochastic variable. Then your own observation that $|E[Y\vert \mathcal{G}] | \leq E[|Y| \vert \mathcal{G}]$ pretty much seals the deal.

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.