geoBayes (version 0.5.1)

rsglmm: Simulation from a spatial model

Description

Simulate from a variety of spatial models.

Usage

rsglmm(n, formula, family = "gaussian", data, weights, subset, atsample, beta,
  linkp, phi, omg, kappa, ssq, corrfcn = "matern", longlat = FALSE,
  dispersion = 1, returnGRF = FALSE, warndisp = TRUE)
rstrga(n, formula, data, weights, subset, atsample, beta, linkp, phi, omg,
  kappa, ssq, corrfcn = "matern", longlat = FALSE, dispersion = 1,
  returnGRF = FALSE)
rsgrf(n, formula, data, subset, atsample, beta, phi, omg, kappa, ssq,
  corrfcn = "matern", longlat = FALSE)

Arguments

n

The number of instances to simulate

formula

A representation of the model in the form response ~ terms. The LHS can be omitted. If the LHS exists, it can be of the form y, y|z, or sums of terms at either side of the | to specify the names of the variables to include in the data frame.

family

The distribution of the data to simulate from.

data

An optional data frame containing the variables in the model.

weights

An optional vector of weights. Number of replicated samples for Gaussian and gamma, number of trials for binomial, time length for Poisson.

subset

An optional set of indices. Simulations will be provided for those locations only.

atsample

A formula of the form ~ Xcoord + Ycoord specifying the sampled locations.

beta

A vector of the regressor coefficents to use.

linkp

The link function parameter.

phi

The spatial range parameter.

omg

The relative nugget parameter.

kappa

The spatial smoothness parameter.

ssq

The partial sill parameter.

corrfcn

The correlation function to use.

longlat

How to compute the distance between locations. If FALSE, Euclidean distance, if TRUE Great Circle distance. See spDists.

dispersion

The fixed dispersion parameter. When this is not 1 and the sample is from a binomial or a Poisson distribution, an approximate sample is returned.

returnGRF

Whether to return the simulate Gaussian random field as well.

warndisp

Whether to warn when sampling from a quasi distribution. This is the case for binomial and Poisson when the dispersion is not one.

Value

A data frame containing the predictors, sampling locations, optional weights, and samples.

Details

The spatial Gaussian random field is simulated using the Cholesky decomposition of the covariance matrix.

The sample from a quasi distribution uses a hack which matches the mean and the variance of the distribution. See the source code for details.

Examples

# NOT RUN {
n <- 100
beta <- c(-2, 1)
phi <- .2
omg <- .3
linkp <- 0
ssq <- 1
l <- rep(10, n)
corrf <- "matern"
kappa <- .5
family <- "poisson"
Xcoord <- runif(n)
Ycoord <- runif(n)
f <- Xcoord + Ycoord
formula <- y|z ~ f
mydata <- rsglmm(1, formula, family, weights = l,
                 atsample = ~ Xcoord + Ycoord, beta = beta, linkp = linkp,
                 phi = phi, omg = omg, kappa = kappa, ssq = ssq,
                 corrfcn = corrf, returnGRF = TRUE)
# }