Definition:Conditional Expectation
Definition
General Case
Let $\struct {\Omega, \Sigma, \Pr}$ be a probability space.
Let $X$ be an integrable random variable on $\struct {\Omega, \Sigma, \Pr}$.
Conditioned on $\sigma$-Algebra
Let $\GG \subseteq \Sigma$ be a sub-$\sigma$-algebra of $\Sigma$.
We say that $Z$ is a version of the conditional expectation of $X$ given $\GG$, or version of $\expect {X \mid \GG}$ if and only if:
- $(1): \quad \expect {\cmod Z} < \infty$
- $(2): \quad$ $Z$ is $\GG$-measurable
- $(3): \quad \ds \forall G \in \GG: \int_G Z \rd \Pr = \int_G X \rd \Pr$
From Existence and Essential Uniqueness of Conditional Expectation Conditioned on Sigma-Algebra, any two versions of the conditional expectation of $X$ given $\GG$ agree almost surely, so we write:
- $Z = \expect {X \mid \GG}$
in the sense of almost-sure equality.
Conditioned on Set of Random Variables
Let $\SS$ be a set of real-valued random variables on $\struct {\Omega, \Sigma, \Pr}$.
Then we define the conditional expectation of $X$ given $\SS$:
- $\expect {X \mid \SS} = \expect {X \mid \map \sigma \SS}$
where:
- $\map \sigma \SS$ denotes the $\sigma$-algebra generated by $\SS$
- $\expect {X \mid \map \sigma \SS}$ denotes the conditional expectation of $X$ given $\map \sigma \SS$
- $=$ is understood to mean almost-sure equality.
If $\SS$ is countable set, say $\SS = \set {X_n : n \in \N} = \set {X_1, X_2, \ldots}$, we may write:
- $\expect {X \mid \SS} = \expect {X \mid X_1, X_2, \ldots}$
Conditioned on Event
Let $A \in \Sigma$.
Then we define the conditional expectation of $X$ given $A$:
- $\expect {X \mid A} = \expect {X \mid \map \sigma A}$
where:
- $\map \sigma A$ denotes the $\sigma$-algebra generated by $A$
- $\expect {X \mid \map \sigma A}$ denotes the conditional expectation of $X$ given $\map \sigma A$
- $=$ is understood to mean almost-sure equality.
Discrete Case
Let $\struct {\Omega, \Sigma, \Pr}$ be a probability space.
Let $X$ be a discrete random variable on $\struct {\Omega, \Sigma, \Pr}$.
Let $B$ be an event in $\struct {\Omega, \Sigma, \Pr}$ such that $\map \Pr B > 0$.
The conditional expectation of $X$ given $B$ is written $\expect {X \mid B}$ and defined as:
- $\expect {X \mid B} = \ds \sum_{x \mathop \in \image X} x \condprob {X = x} B$
where:
- $\condprob {X = x} B$ denotes the conditional probability that $X = x$ given $B$
whenever this sum converges absolutely.
Also known as
Conditional expectation is also known as conditional mean.
Also see
- Results about conditional expectation can be found here.
Sources
- 1998: David Nelson: The Penguin Dictionary of Mathematics (2nd ed.) ... (previous) ... (next): regression
- 2008: David Nelson: The Penguin Dictionary of Mathematics (4th ed.) ... (previous) ... (next): regression