= Gaussian conditional expectation
{c}
{title2=$\mathbb E[Y\mid S]$}
Let the joint <random vector> $(Y,S)$ have a <multivariate normal distribution>, with <expectations> $\mu_Y,\mu_S$, <covariance matrices> $\Sigma_Y,V$ and cross-<covariance matrix> $C=\operatorname{cov}(Y,S)$. If $V$ is invertible, the <conditional expectation> is the affine <function>
$$
\mathbb E[Y\mid S]=\mu_Y+CV^{-1}(S-\mu_S),
$$
and the <conditional covariance> <matrix> is $\Sigma_Y-CV^{-1}C^T$.
To prove both assertions, define $R=Y-\mu_Y-CV^{-1}(S-\mu_S)$. Direct <covariance> calculation gives $\operatorname{cov}(R,S)=0$. Since $(R,S)$ has a <multivariate normal distribution>, this zero cross-<covariance> implies <independence>. The decomposition of $Y$ into its affine part and $R$ therefore gives the <conditional expectation> and the remaining <covariance matrix>. This is the <conditional multivariate normal distribution> calculation.
If $V$ is singular, replace $V^{-1}$ by its <Moore-Penrose inverse> $V^+$. The formulas hold for $S$ in its affine <probability support>, $\mu_S+\operatorname{range}V$; values away from that <probability support> do not determine a unique version of the <conditional expectation>. For example, if $S=(Z,Z)^T$ and $Y=Z$ for a standard <normal random variable> $Z$, then $V$ is the all-ones $2\times2$ <matrix>. Although $V^{-1}$ does not exist, $CV^+S=(S_1+S_2)/2=Z$ on the <probability support> $S_1=S_2$.
Back to article page