posterior.sample: Generate samples from an approximated posterior of a fitted...

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

Description

This function generate samples from an approximated posterior of a fitted model, ie an inla-object

Usage

     inla.posterior.sample(n = 1L, result, hyper.user.scale = TRUE, use.improved.mean = TRUE)
 

Arguments

n	Number of samples.
result	The inla-object, ie the output from an inla-call. The inla-object must be created with control.compute=list(config=TRUE).
hyper.user.scale	Logical. If TRUE then values of the hyperparameters are given in the user scale (for example precision). If FALSE then values of the hyperparameters are given in the internal representation (for example log(precision)).
use.improved.mean	Logical. If TRUE then use the marginal mean values when constructing samples. If FALSE then use the mean in the Gaussian approximations.

Value

A list of the samples, where each sample is a list with names hyperpar and latent, and with their marginal densities in logdens$hyperpar and logdens$latent and the joint density is in logdens$joint. THIS IS AN EXPERIMENTAL FUNCTION AND CHANGES MAY APPEAR AT ANY TIME!

Author(s)

Havard Rue [e-mail]

Examples

  r = inla(y ~ 1 ,data = data.frame(y=rnorm(1)), control.compute = list(config=TRUE))
  samples = inla.posterior.sample(2,r)

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