If \(X\) is an integrable random variable on \((\Omega, \mathcal{F}, P)\), and \(\mathcal{G}\) is a sub-\(\sigma\)-field of \(\mathcal{F}\), then there exists a \(\mathcal{G}\)-measurable random variable \(Y\), unique up to almost sure equality, which satisfies \[
\int_A Y dP = \int_A X dP \mbox{ }\mbox{ }\mbox{ for all }A\in \mathcal{G}.
\]
Let \(X\) be an integrable random variable on a probability space \((\Omega, \mathcal{F},P)\) and let \(\mathcal{G}\) be a \(\sigma\)-filed with \(\mathcal{G}\subset \mathcal{F}\). Then, the conditional expectation of \(X\) given \(\mathcal{G}\), denoted \(E(X|\mathcal{G})\) is defined to be that random variable, unique up to almost sure equality, satsfying two conditions:
\(E(X|\mathcal{G})\) is \(\mathcal{G}\)-measurable, and
\(\int_A E(X|\mathcal{G})=\int_A X dP\) for all \(A\in \mathcal{G}\).
Any random variable \(Y\) satisfying 1,2 is said to be a version of \(E(X|\mathcal{G})\).
우변에서의 괄호 안의 값은 상수이고 두개의 indicator function들은 모두 \(\mathcal{G}\)-measurable이므로 linear combination of \(\mathcal{G}\)-measurable function \(\implies\) \(\mathcal{G}\)-measurable이다.
Note: \[
\int_A E(X|\mathcal{G}) dP = \left( \frac{1}{P(B)}\int_B X \mbox{ }dP\right)P(A\cap B)+ \left( \frac{1}{P(B^c)}\int_{B^c} X \mbox{ }dP\right)P(A\cap B^c)=\int_A\mbox{ } XdP \mbox{ }\mbox{ }\mbox{ }\mbox{ for all } A\in \mathcal{G},
\] 왜냐하면 \(A\)가 \(\phi\)면 당연히 \(0=0\)이므로 성립하고, \(\Omega\)면
\[
\int_\Omega E(X|\mathcal{G})dP=\left( \frac{1}{P(B)}\int_B X \mbox{ }dP\right)P(B) + \left( \frac{1}{P(B^c)}\int_{B^c} X \mbox{ }dP\right)P({B^c})=\int_\Omega X dP
\] 이다. 또한 \(A=B\)이면 \(\int_B E(X|\mathcal{G})dP=\left( \frac{1}{P(B)}\int_B X \mbox{ }dP\right)P(B) =\int_B X dP\)이고 \(B^c\)도 이와 같이 할 수 있다.
If \(\mathcal{G}\) is a sub-\(\sigma\)-field of \(\mathcal{F}\) and \(A\in \mathcal{F}\), then the conditional probability of \(A\) given \(\mathcal{G}\), denoted \(P(A|\mathcal{G})\) is defined to be \(P(A|\mathcal{G})=E(I_A|\mathcal{G})\).
\[
P(A|\mathcal{G})= E(I_A|\mathcal{G})(\omega)=\begin{cases}\frac{P(A\cap B)}{P(B)},& \mbox{ if }\omega\in B,\\ \frac{P(A\cap B^c)}{P(B^c)},& \mbox{ if }\omega\in B^c.\end{cases}
\]
즉, \(\mathcal{G}\)의 set \(B\)에 대해서는 \(P(A|B)=\frac{P(A\cap B)}{P(B)}\), \(B^c\)에 대해서는 \(P(A|B^c)= \frac{P(A\cap B^c)}{P(B^c)}\)라는 표현이 \(P(A|\mathcal{G})\)로부터 비롯된다.
If \(X\) is an integrable r.v and \(Y\) is a random vector, both defined on a common probability space \((\Omega, \mathcal{F}, P)\), then the conditional expectation of \(X\) given \(Y\), denoted \(E(X|Y)\) is defined to be \(E[X|\sigma(Y)]\).
Similarly, if \(\{Y_t,t\in T\}\) is any collection of r.vs defined on \((\Omega, \mathcal{F},P)\), then the conditional expectation of \(X\) given \(\{Y_t,t\in T\}\), denoted \(E(X|Y_t,t\in T)\) is defined to be \(E[X|\sigma(Y_t,t\in T)]\).
우선 \(\sigma(Y)\in\mathcal{F}\)이다.
두 변수에 대한 조건부 기댓값 \(E(X|Y)\)는 \(Y\)가 \(\mathcal{F}\)-measurable이기 때문에, \(Y^{-1}(H)\in \sigma(Y)\) for all \(H\in \mathcal{R}\)이다. 즉 \(E(X|Y)\)는 모든 event \(A\in \sigma(Y)\)에 대한 조건부 기댓값을 의미하므로 \(E(X|\sigma(Y))\)와 같다.