lattice2node: Functions to define mapping between a lattice and nodes

In andrewzm/INLA: Functions which allow to perform full Bayesian analysis of latent Gaussian models using Integrated Nested Laplace Approximaxion

Description

These functions define mapping in between two-dimensional indices on a lattice and the one-dimensional node representation used in inla.

The mapping from node to lattice follows the default R behaviour (which is column based storage), and as.vector(A) and matrix(a, nrow, ncol) can be used instead of inla.matrix2vector and inla.vector2matrix.

Usage

inla.lattice2node.mapping(nrow, ncol)
inla.node2lattice.mapping(nrow, ncol)
inla.lattice2node(irow, icol, nrow, ncol)
inla.node2lattice(node, nrow, ncol)
inla.matrix2vector(a.matrix)
inla.vector2matrix(a.vector, nrow, ncol)

Arguments

nrow	Number of rows in the lattice.
ncol	Number of columns in the lattice.
irow	Lattice row index, between 1 and nrow
icol	Lattice column index, between 1 and ncol
node	The node index, between 1 and ncol*nrow
a.matrix	is a matrix to be mapped to a vector using internal representation defined by inla.lattice2node
a.vector	is a vector to be mapped into a matrix using the internal representation defined by inla.node2lattice

Value

inla.lattice2node.mapping returns the hole mapping as a matrix, and inla.node2lattice.mapping returns the hole mapping as list(irow=..., icol=...). inla.lattice2node and inla.node2lattice provide the mapping for a given set of lattice indices and nodes. inla.matrix2vector provide the mapped vector from a matrix, and inla.vector2matrix provide the inverse mapped matrix from vector.

Author(s)

Havard Rue [e-mail]

See Also

inla

Examples

## write out the mapping using the two alternatives
nrow = 2
ncol = 3
mapping = inla.lattice2node.mapping(nrow,ncol)

for (i in 1:nrow){
    for(j in 1:ncol){
        print(paste("Alt.1: lattice index [", i,",", j,"] corresponds",
                    "to node [", mapping[i,j],"]", sep=""))
    }
}

for (i in 1:nrow){
    for(j in 1:ncol){
        print(paste("Alt.2: lattice index [", i,",", j,"] corresponds to node [",
                    inla.lattice2node(i,j,nrow,ncol), "]", sep=""))
    }
}

inv.mapping = inla.node2lattice.mapping(nrow,ncol)
for(node in 1:(nrow*ncol))
   print(paste("Alt.1: node [", node, "] corresponds to lattice index [",
               inv.mapping$irow[node], ",",
               inv.mapping$icol[node],"]", sep=""))

for(node in 1:(nrow*ncol))
   print(paste("Alt.2: node [", node, "] corresponds to lattice index [",
               inla.node2lattice(node,nrow,ncol)$irow[1], ",",
               inla.node2lattice(node,nrow,ncol)$icol[1],"]", sep=""))

## apply the mapping from matrix to vector and back
n = nrow*ncol
z = matrix(1:n,nrow,ncol)
z.vector = inla.matrix2vector(z)  # as.vector(z) could also be used
print(mapping)
print(z)
print(z.vector)

## the vector2matrix is the inverse, and should give us the z-matrix
## back. matrix(z.vector, nrow, ncol) could also be used here.
z.matrix = inla.vector2matrix(z.vector, nrow, ncol) 
print(z.matrix)

andrewzm/INLA documentation built on May 10, 2019, 11:12 a.m.