Let's say my data of dimension $3$ has a multivariate normal distribution with known mean vector $\mu$ and covariance matrix $\Sigma$

I then observe a sample from the distribution but with $1$ value missing $[ ?, x_2, x_3 ]$

How can I find the most likely value of $x_1$ given $x_2$ and $x_3$?


Condition on $(x_2, x_3)$, $P(x_1|x_2,x_3)$ is still a Gaussian distribution.

Its mean (which coincides with its mode) is

$$\mu_1+\begin{bmatrix}\Sigma_{12} & \Sigma_{12} \end{bmatrix}\begin{bmatrix}\Sigma_{22} & \Sigma_{23} \\ \Sigma_{32} & \Sigma_{33}\end{bmatrix}^{-1}\begin{bmatrix} x_2 - \mu_2 \\ x_3-\mu_3 \end{bmatrix}. $$


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.