Conditional Expectation Conditioned on Event of Non-Zero Probability
Theorem
Let $\struct {\Omega, \Sigma, \Pr}$ be a probability space.
Let $X$ be an integrable random variable on $\struct {\Omega, \Sigma, \Pr}$.
Let $B \in \Sigma$ be an event with:
- $\map \Pr B > 0$
Let:
- $\GG = \map \sigma B = \set {\O, B, B^c, \Omega}$
where $\map \sigma B$ is the $\sigma$-algebra generated by $B$.
Let:
- $\ds \alpha = \frac {\expect {X \cdot 1_B} } {\map \Pr B}$
and:
- $\ds \beta = \frac {\expect {X \cdot 1_{B^c} } } {\map \Pr {B^c} }$
Let $\expect {X \mid \GG}$ be a version of the conditional expectation of $X$ given $\GG$.
Then:
- $\ds \expect {X \mid \GG} = \alpha \cdot 1_B + \beta \cdot 1_{B^c}$ almost everywhere.
Proof
We show that:
- $\ds Z = \alpha \cdot 1_B + \beta \cdot 1_{B^c}$ is a version of the conditional expectation of $X$ given $\GG$.
We will then be done by Existence and Essential Uniqueness of Conditional Expectation Conditioned on Sigma-Algebra.
From Characteristic Function Measurable iff Set Measurable, $1_B$ and $1_{B^c}$ are both measurable functions.
From Pointwise Product of Measurable Functions is Measurable and Pointwise Sum of Measurable Functions is Measurable, we therefore have that $Z$ is a random variable.
So we now verify that the expectations take the correct values.
From Integral of Integrable Function over Null Set, we have:
- $\expect {Z \cdot 1_\O} = 0 = \expect {X \cdot 1_\O}$
We also have:
\(\ds \expect {Z \cdot 1_B}\) | \(=\) | \(\ds \expect {\alpha \cdot 1_B + \beta \cdot 1_B \cdot 1_{B^c} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \expect {\alpha \cdot 1_B}\) | Characteristic Function of Intersection | |||||||||||
\(\ds \) | \(=\) | \(\ds \alpha \map \Pr B\) | Expectation is Linear: General Case, Integral of Characteristic Function | |||||||||||
\(\ds \) | \(=\) | \(\ds \expect {X \cdot 1_B}\) |
and:
\(\ds \expect {Z \cdot 1_{B^c} }\) | \(=\) | \(\ds \expect {\alpha \cdot 1_B \cdot 1_{B^c} + \beta \cdot 1_{B^c} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \expect {\beta \cdot 1_{B^c} }\) | Characteristic Function of Intersection | |||||||||||
\(\ds \) | \(=\) | \(\ds \beta \map \Pr {B^c}\) | Expectation is Linear: General Case, Integral of Characteristic Function | |||||||||||
\(\ds \) | \(=\) | \(\ds \expect {X \cdot 1_{B^c} }\) |
Finally, we have:
\(\ds \expect Z\) | \(=\) | \(\ds \expect {\alpha \cdot 1_B + \beta \cdot 1_{B^c} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \alpha \map \Pr B + \beta \map \Pr {B^c}\) | Expectation is Linear: General Case, Integral of Characteristic Function | |||||||||||
\(\ds \) | \(=\) | \(\ds \expect {X \cdot 1_B} + \expect {X \cdot 1_{B^c} }\) | ||||||||||||
\(\ds \) | \(=\) | \(\ds \expect {X \cdot \paren {1_B + 1_{B^c} } }\) | Expectation is Linear: General Case | |||||||||||
\(\ds \) | \(=\) | \(\ds \expect X\) | Characteristic Function of Disjoint Union |
$\blacksquare$