pc-multvar: Multivariate PC priors

In inbo/INLA: Full Bayesian Analysis of Latent Gaussian Models using Integrated Nested Laplace Approximations

Description

Functions to evaluate and simulate from multivariate PC priors: The simplex and sphere case

Usage

 inla.pc.multvar.h.default(x, inverse = FALSE, derivative = FALSE)
 inla.pc.multvar.simplex.r(n = NULL, lambda = 1, h = inla.pc.multvar.h.default, b = NULL)
 inla.pc.multvar.simplex.d(x = NULL, lambda = 1, log = FALSE, h = inla.pc.multvar.h.default, b = NULL)
 inla.pc.multvar.sphere.r(n = NULL, lambda = 1, h = inla.pc.multvar.h.default, H = NULL)
 inla.pc.multvar.sphere.d(x = NULL, lambda = 1, log = FALSE, h = inla.pc.multvar.h.default, H = NULL)
 

Arguments

x	Samples to evaluate. If input is a matrix then each row is a sample. If input is a vector then this is the sample.
inverse	Compute the inverse of the h()-function.
derivative	Compute the derivative of the h()-function. (derivative of the inverse function is not used).
n	Number of samples to generate.
lambda	The lambda-parameter in the PC-prior.
log	Evaluate the density in log-scale or ordinary scale.
h	The h()-function, defaults to inla.pc.multvar.h.default. See that code for an example of how to write a user-spesific function.
b	The b-vector (gradient) in the expression for the simplex option, d(xi) = h(b^T xi)
H	The H(essian)-matrix in the expression for the sphere option, d(xi) = h(1/2 *xi^T H xi). If H is a vector, then it is interpreted as the diagonal of a (sparse) diagonal matrix.

Details

These functions implements multivariate PC-priors of the simplex and sphere type.

Value

inla.pc.multvar.simplex.r generate samples from the simplex case, and inla.pc.multvar.simplex.d evaluate the density. inla.pc.multvar.sphere.r generate samples from the sphere case, and inla.pc.multvar.sphere.d evaluate the density. inla.pc.multvar.h.default implements the default h()-function and illustrate how to code your own spesific one, if needed.

Author(s)

Havard Rue [e-mail]

inbo/INLA documentation built on Dec. 6, 2019, 9:51 a.m.