CAR.simLM: Simulate samples from a CAR model.

In mclcar: Estimating Conditional Auto-Regressive (CAR) Models using Monte Carlo Likelihood Methods

Description

This help page documents several functions for simulate samples from different CAR models.

Usage

CAR.simTorus(n1, n2, rho, prec)
CAR.simWmat(rho, prec, W)
CAR.simLM(pars, data)
CAR.simGLM(method = c("binom", "poisson"), W, n = NULL, pars, Xs = NULL,
n.trial = 1)

Arguments

n1, n2, n	n = c(n1, n2), the size of the torus
rho	the spatial coefficient in the CAR precision matrix (inverse of the covariance matrix)
prec	the precision in the CAR precision matrix
W	the spatial weight matrix
pars	parameter values of the direct CAR model
data	a data object same as described in loglik.dCAR
method	a character equal to either "binom" or "poisson" that indicate the distribution of the sample to be simulated.
Xs	the covariates to be used in the glm
n.trial	the number of trials in each unit of the binomial samples

Value

CAR.simTorus	returns a list containing the spatial weight matrix W and the simulated CAR samples X;
CAR.simWmat,CAR.simLM	returns an array of the simulated CAR samples;
CAR.simGLM	returns a list containing: rho, sigma, beta, the parameter values used to generate the samples y, the simulated data covX, the covariates W, the spatial weight matrix Z.true, the simulated CAR variables eta, the simulated glm linear response Emean, the simulated glm mean n.trial, the number of trials in each unit for binomial samples

Author(s)

Zhe Sha [e-mail]

See Also

mcl.glm, mcl.dCAR, mcl.prep.dCAR, mcl.prep.glm

Examples

## Simulate CAR data on a torus
set.seed(33)
n.torus <- 10
rho <- 0.2
sigma <- 1.5
prec <- 1/sigma
beta <- c(1, 1)
XX <- cbind(rep(1, n.torus^2), sample(log(1:n.torus^2)/5))
mydata1 <- CAR.simTorus(n1 = n.torus, n2 = n.torus, rho = rho, prec = prec)

## Simulate CAR data for a given spatial weight matrix
Wmat <- mydata1$W
mydata2 <- CAR.simWmat(rho = rho, prec = prec, W = Wmat)

## Simualte data from a linear model with CAR error
y <- XX %*% beta + mydata1$X
mydata1$data.vec <- data.frame(y=y, XX[,-1])
mydata3 <- CAR.simLM(pars = c(0.1, 1, 2, 0.5), data = mydata1)

## Simulate Binomial data with CAR latent variables
mydata4 <- CAR.simGLM(method="binom", n=c(10,10), pars = c(rho, sigma,
                      beta), Xs=XX, n.trial = 5)

Example output

mclcar documentation built on Jan. 8, 2022, 5:07 p.m.