2
$\begingroup$

The conditional expectation $\mathbb{E}[\,X\,|\,\mathcal{G}\,]$ of a random variable $X\in\mathcal{L}^{1}$ on some probability space $(\Omega,\mathcal{F},\mathbb{P})$ (where $\mathcal{G}\subset\mathcal{F}$) always exists and is unique up to sets of measure zero.

I am wondering what this last part really means. Here is my understanding: $\mathbb{E}[\,X\,|\,\mathcal{G}\,]$ is unique on all $A\in\mathcal{G}$ with $\mathbb{P}(A)=1$. Is this correct? Or does uniqueness only hold on (at least) one $A\in\mathcal{G}$ with full measure?

On a similar note: the (conditional) dominated convergence theorem states that $$\mathbb{E}[\,X_{n}\,|\,\mathcal{G}\,] \quad\rightarrow\quad \mathbb{E}[\,X\,|\,\mathcal{G}\,]\qquad\mathbb{P}\text{-a.s.}$$ for $X_{n}\rightarrow X$ $\mathbb{P}$-a.s. with $X_{n}\leq Y$, $Y\in\mathcal{L}^1$.

Again: Does this mean, that I can choose some set $A\in\mathcal{G}$ with $\mathbb{P}(A)=1$, such that $$\mathbb{E}[\,X_{n}\,|\,\mathcal{G}\,](\omega) \quad\rightarrow\quad \mathbb{E}[\,X\,|\,\mathcal{G}\,](\omega)$$ for all $\omega\in A$?

Your help is highly appreciated! Thank you.

$\endgroup$
3
$\begingroup$

Remember that $\mathbb{E}\left[X\mid\mathcal{G}\right]$ is defined to be a $\mathcal{G}$-measurable random variable such that $$\int_A X d\mathbb{P}=\int_A\mathbb{E}\left[X\mid\mathcal{G}\right]d\mathbb{P}$$ for all $A\in\mathcal{G}$. If one changes the value of $\mathbb{E}\left[X\mid\mathcal{G}\right]$ at a set of measure zero, the value of the integral will not change. So saying that $\mathbb{E}\left[X\mid\mathcal{G}\right]$ is unique up to sets of measure zero means that each two random variables which satisfy the above requirements will only differ on a set of measure zero.

As for your second question -- what you wrote is correct.

$\endgroup$
  • $\begingroup$ Thank you for the clarification! $\endgroup$ – Mark Jun 2 '17 at 14:59

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.