bw.pcf.Rd
\name{bw.pcf}
\alias{bw.pcf}
\title{
Cross Validated Bandwidth Selection for Pair Correlation Function
}
\description{
Uses composite likelihood or generalized least squares
cross-validation to select a smoothing bandwidth
for the kernel estimation of pair correlation function.
}
\usage{
bw.pcf(X, rmax=NULL, lambda=NULL, divisor="r",
kernel="epanechnikov", nr=10000, bias.correct=TRUE,
cv.method=c("compLik", "leastSQ"), simple=TRUE, srange=NULL, \dots)
}
\arguments{
\item{X}{
A point pattern (object of class \code{"ppp"}).
}
\item{rmax}{
Numeric. Maximum value of the spatial lag distance \eqn{r}
for which \eqn{g(r)} should be evaluated.
}
\item{lambda}{
Optional.
Values of the estimated intensity function.
A vector giving the intensity values
at the points of the pattern \code{X}.
}
\item{divisor}{
Choice of divisor in the estimation formula:
either \code{"r"} (the default) or \code{"d"}.
See \code{pcf.ppp}.
}
\item{kernel}{
Choice of smoothing kernel, passed to \code{density};
see \code{\link{pcf}} and \code{\link{pcfinhom}}.
}
\item{nr}{
Integer. Number of subintervals for discretization of
[0, rmax] to use in computing numerical integrals.
}
\item{bias.correct}{
Logical. Whether to use bias corrected version of the kernel
estimate. See Details.
}
\item{cv.method}{
Choice of cross validation method: either
\code{"compLik"} or \code{"leastSQ"} (partially matched).
}
\item{simple}{
Logical. Whether to use simple removal of spatial lag
distances. See Details.
}
\item{srange}{
Optional. Numeric vector of length 2 giving the range of
bandwidth values that should be searched to find the optimum
bandwidth.
}
\item{\dots}{
Other arguments, passed to \code{\link{pcf}} or
\code{\link{pcfinhom}}.
}
}
\details{
This function selects an appropriate bandwidth \code{bw}
for the kernel estimator of the pair correlation function
of a point process intensity computed by \code{\link{pcf.ppp}}
(homogeneous case) or \code{\link{pcfinhom}}
(inhomogeneous case).
With \code{cv.method="leastSQ"}, the bandwidth
\eqn{h} is chosen to minimise an unbiased
estimate of the integrated mean-square error criterion
\eqn{M(h)} defined in equation (4) in Guan (2007a).
With \code{cv.method="compLik"}, the bandwidth
\eqn{h} is chosen to maximise a likelihood
cross-validation criterion \eqn{CV(h)} defined in
equation (6) of Guan (2007b).
\deqn{
M(b) = \frac{\mbox{MSE}(\sigma)}{\lambda^2} - g(0)
}{
M(b) = \int_{0}^{rmax} \hat{g}^2(r;b) r dr - \sum_{u,v}
}
The result is a numerical value giving the selected bandwidth.
}
\section{Definition of bandwidth}{
The bandwidth \code{bw} returned by \code{bw.pcf}
corresponds to the standard deviation of the smoothoing
kernel. As mentioned in the documentation of
\code{\link{density.default}} and \code{\link{pcf.ppp}},
this differs from the scale parameter \code{h} of
the smoothing kernel which is often considered in the
literature as the bandwidth of the kernel function.
For example for the Epanechnikov kernel, \code{bw=h/sqrt(h)}.
}
\value{
A numerical value giving the selected bandwidth.
The result also belongs to the class \code{"bw.optim"}
which can be plotted.
}
\seealso{
\code{\link{pcf.ppp}},
\code{\link{pcfinhom}}
}
\examples{
b <- bw.pcf(redwood)
plot(pcf(redwood, bw=b))
}
\references{
Guan, Y. (2007a).
A composite likelihood cross-validation approach in selecting
bandwidth for the estimation of the pair correlation function.
\emph{Scandinavian Journal of Statistics},
\bold{34}(2), 336--346.
Guan, Y. (2007b).
A least-squares cross-validation bandwidth selection approach
in pair correlation function estimations.
\emph{Statistics & Probability Letters},
\bold{77}(18), 1722--1729.
}
\author{
Rasmus Waagepetersen and Abdollah Jalilian.
Adapted for \pkg{spatstat} by \spatstatAuthors.
}
\keyword{spatial}
\keyword{methods}
\keyword{smooth}