| scale.model {INLA} | R Documentation |
Scale an intrinsic GMRF model
Description
This function scales an intrinsic GMRF model so the geometric mean of the marginal variances is one
Usage
inla.scale.model.internal(Q, constr = NULL, eps = sqrt(.Machine$double.eps))
inla.scale.model(Q, constr = NULL, eps = sqrt(.Machine$double.eps))
Arguments
Q |
A SPD matrix, either as a (dense) matrix or |
constr |
Linear constraints spanning the null-space of |
eps |
A small constant added to the diagonal of |
Value
inla.scale.model returns a sparseMatrix of type
dgTMatrix scaled so the geometric mean of the marginal variances (of
the possible non-singular part of Q) is one, for each connected
component of the matrix.
Author(s)
Havard Rue hrue@r-inla.org
Examples
## Q is singular
data(Germany)
g = system.file("demodata/germany.graph", package="INLA")
Q = -inla.graph2matrix(g)
diag(Q) = 0
diag(Q) = -rowSums(Q)
n = dim(Q)[1]
Q.scaled = inla.scale.model(Q, constr = list(A = matrix(1, 1, n), e=0))
print(diag(MASS::ginv(as.matrix(Q.scaled))))
## Q is singular with 3 connected components
g = inla.read.graph("6 1 2 2 3 2 2 1 3 3 2 1 2 4 1 5 5 1 4 6 0")
print(paste("Number of connected components", g$cc$n))
Q = -inla.graph2matrix(g)
diag(Q) = 0
diag(Q) = -rowSums(Q)
n = dim(Q)[1]
Q.scaled = inla.scale.model(Q, constr = list(A = matrix(1, 1, n), e=0))
print(diag(MASS::ginv(as.matrix(Q.scaled))))
## Q is non-singular with 3 connected components. no constraints needed
diag(Q) = diag(Q) + 1
Q.scaled = inla.scale.model(Q)
print(diag(MASS::ginv(as.matrix(Q.scaled))))
[Package INLA version 25.06.13 Index]