0
$\begingroup$

I'm currently going through a book on measure-theoretic probability. In it, we defined conditional expectation:

Let X belong to $M^+(Omega, F)$ and P be a probability measure on F. For each sub-sigma-field G of F, the conditional expectation$ P(X|G)$ is the random variable $X_G$ in $M^+(X,G)$ for which $P(gX) = P(gX_G)$.

However, there is a small disclaimer: be careful to check that $X_G$ is G-measurable. I'm confused about how this can be a problem, since isn't $X_G\in M^+(X,G)$? Doesn't this by definition make it $G-$measurable?

$\endgroup$
1
  • 2
    $\begingroup$ \Omega produces $\Omega$ $\endgroup$ – parsiad Apr 3 '19 at 0:40
0
$\begingroup$

They are just saying that if you want to prove that $Y$ is $P(X|G)$ you have to verify two facts: $Y \in M^{+}(X,G)$ and the condition $P(gX)=P(gY)$. A common mistake it to ignore the first part.

$\endgroup$
2
  • $\begingroup$ Oh I see.. It says "check that X_G is G-measurable. Otherwise you might be tempted to leap to the conclusion that X_G = X". But we don't know that X is G-measurable, making this equality perhaps false $\endgroup$ – swedishfished Apr 3 '19 at 0:47
  • $\begingroup$ @swedishfished Eaxclty. That is why they are issuing a word of caution. $\endgroup$ – Kavi Rama Murthy Apr 3 '19 at 0:48

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.