As quick_mvnorm(), but taking precomputed Cholesky factors so repeated
draws (e.g. the perturbation draws in gp_predict()) skip the
factorisation cost.
Usage
quick_mvnorm_chol(space_chol, time_chol)
Arguments
- space_chol
Upper-triangular Cholesky factor of the space kernel
matrix, as returned by chol(). Passing the lower-triangular factor
gives silently wrong draws.
- time_chol
Upper-triangular Cholesky factor of the time kernel
matrix, as returned by chol().
Value
A numeric vector of length nrow(space_chol) * nrow(time_chol),
ordered site by site with time varying fastest (matching
kronecker(space, time)).