sample.geodata: Sampling from geodata objects

Description

This functions facilitates extracting samples from geodata objects.

Usage

sample.geodata(x, size, replace = FALSE, prob = NULL, coef.logCox,
               external)

Arguments

an object of the class geodata.

size

non-negative integer giving the number of items to choose.

replace

Should sampling be with replacement?

prob

A vector of probability weights for obtaining the elements of the data points being sampled.

coef.logCox

optional. A scalar with the coeficient for the log-Cox process. See DETAILS below.

external

numeric values of a random field to be used in the log-Cox inhomogeneous poisson process.

Value

a list which is an object of the class geodata.

Details

If prob=NULL and the argument coef.logCox, is provided, sampling follows a log-Cox proccess, i.e. the probability of each point being sampled is proportional to: $$exp(b Y(x))$$ with $b$ given by the value passed to the argument coef.logCox and $Y(x)$ taking values passed to the argument external or, if this is missing, the element data of the geodata object. Therefore, the latter generates a preferential sampling.

Examples

Run this code

# NOT RUN {
par(mfrow=c(1,2))
S1 <- grf(2500,  grid="reg", cov.pars=c(1, .23))
image(S1, col=gray(seq(0.9,0.1,l=100)))
y1 <- sample.geodata(S1, 80)
points(y1$coords, pch=19)
## Now a preferential sampling
y2 <- sample.geodata(S1, 80, coef=1.3)
## which is equivalent topps
## y2 <- sample.geodata(S1, 80, prob=exp(1.3*S1$data))
points(y2$coords, pch=19, col=2)
## and now a clustered (but not preferential)
S2 <- grf(2500,  grid="reg", cov.pars=c(1, .23))
y3 <- sample.geodata(S1, 80, prob=exp(1.3*S2$data))
## which is equivalent to
## points(y3$coords, pch=19, col=4)
image(S2, col=gray(seq(0.9,0.1,l=100)))
points(y3$coords, pch=19, col=4)
# }