contributed-1: Computes the standard deviation for the structured (random)...

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

Description

Computes the posterior distribution of the standard deviations for the structured (random) effects in an INLA model, starting from the default output based on the precisions

Usage

    inla.contrib.sd(model, nsamples=1000)

Arguments

model	An INLA model, fitted calling the inla()-function. The formula specified for the model should include at least one structured (random) effect in the form f(variable, model="iid").
nsamples	The number of simulations from the posterior distribution of the standard deviations used to compute the summary statistics

Value

inla.contrib.sd returns a matrix samples including the simulated values from the posterior distributions as well as a summary table hyper reporting the mean, sd and 95% credible interval for the posterior distributions of each random effect.

Author(s)

Gianluca Baio [e-mail]

See Also

inla

Examples

# Data generation
n=12
Ntrials = sample(c(80:100), size=n, replace=TRUE)
eta = rnorm(n,0,0.5)
prob = exp(eta)/(1 + exp(eta))
y = rbinom(n, size=Ntrials, prob = prob)
data=data.frame(y=y,z=1:n)
formula=y~f(z,model="iid")
m=inla(formula,data=data,family="binomial",Ntrials=Ntrials)
summary(m)
s=inla.contrib.sd(m)
s$hyper
hist(s$samples,xlab="standard deviation for z",main="")

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