mra_wendland_2d: Code to construct the mutli-resolution sparse basis function...
In BayesMRA: Bayesian Multi-Resolution Gaussian Process Approximations

Description Usage Arguments Value Examples

Code to construct the mutli-resolution sparse basis function representation for fitting spatial processes

mra_wendland_2d(
  locs,
  M = 4,
  n_coarse_grid = 10,
  n_padding = 5L,
  n_neighbors = 68,
  use_spam = TRUE
)

`locs`	The location variables in 2 dimensions over which to construct the basis function representation
`M`	The number of resolutions.
`n_coarse_grid`	The number of basis functions in one direction (e.g. `n_coarse_grid = 10` results in a 10x10 course grid which is further extended by the number of additional padding basis functions given by `n_padding`.
`n_padding`	The number of additional boundary points to add on each boundary. For example, n_padding = 5 will add 5 boundary knots to the both the left and right side of the grid).
`n_neighbors`	The expected number of neighbors for each interior basis function. This determines the basis radius parameter.
`use_spam`	is a boolean flag to determine whether the output is a list of `spam::spam` matrix objects (`use_spam = TRUE`) or a an n x n sparse Matrix of class `Matrix::dgCMatrix` `use_spam = FALSE` (see spam and Matrix packages for details).

A list of objects including the MRA knots locations locs_grid, the Wendland basis representation matrix W at the observed locations, the basis radius radius, the numbers of resolutions M, the number of expected neighbors in the interior of each grid n_neighbors, the number of interior basis functions in one direction n_coarse_grid, the number of additional padding basis functions given by n_padding, and the setting use_spam which determines whether the MRA output uses the spam format.

set.seed(111)
locs <- matrix(runif(20), 10, 2)
MRA <- mra_wendland_2d(locs, M = 2, n_coarse_grid = 4)
## plot the MRA grid at different resolutions
layout(matrix(1:2, 1, 2))
plot(MRA$locs_grid[[1]])
plot(MRA$locs_grid[[2]])