Quick multivariate normal draw over two dimensions

Draws one sample from a zero-mean Gaussian with separable space-time covariance, without ever forming the full matrix. This is equivalent to forming the full spatio-temporal covariance and drawing from the multivariate normal distribution:

Usage

quick_mvnorm(space, time)

Arguments

space

Space kernel matrix.

time

Time kernel matrix.

Value

A numeric vector of length nrow(space) * nrow(time), ordered site by site with time varying fastest (matching kronecker(space, time)).

Details

full_k <- kronecker(space, time)
f <- MASS::mvrnorm(1, rep(0, nrow(space) * nrow(time)), full_k)