\name{localpcf}
\alias{localpcf}
\title{Local pair correlation function}
\description{
  Computes individual contributions to the
  pair correlation function from each data point.
}
\usage{
  localpcf(X, ..., delta=NULL, rmax=NULL, nr=512, stoyan=0.15)
}
\arguments{
  \item{X}{A point pattern (object of class \code{"ppp"}).}
  \item{\dots}{Ignored.}
  \item{delta}{
    Smoothing bandwidth. The halfwidth of the Epanechnikov kernel.
  }
  \item{rmax}{
    Optional. Maximum value of distance \eqn{r} for which
    pair correlation values \eqn{g(r)} should be computed.
  }
  \item{nr}{
    Optional. Number of values of distance \eqn{r} for which
    pair correlation \eqn{g(r)} should be computed.
  }
  \item{stoyan}{
    Optional. The value of the constant \eqn{c} in Stoyan's rule
    of thumb for selecting the smoothing bandwidth \code{delta}.
  }
}
\details{
  \code{localpcf} computes the contribution, from each individual
  data point in a point pattern \code{X}, to the
  empirical pair correlation function of \code{X}.
  These contributions are sometimes known as LISA (local indicator
  of spatial association) functions based on pair correlation.
  
  Given a spatial point pattern \code{X}, the local pcf
  \eqn{g_i(r)}{g[i](r)} associated with the \eqn{i}th point
  in \code{X} is computed by
  \deqn{
    g_i(r) = \frac a {2 \pi n} \sum_j k(d_{i,j} - r) 
  }{
    g[i](r) = (a/(2 * pi * n) * sum[j] k(d[i,j] - r) 
  }
  where the sum is over all points \eqn{j \neq i}{j != i},
  \eqn{a} is the area of the observation window, \eqn{n} is the number
  of points in \code{X}, and \eqn{d_{ij}}{d[i,j]} is the distance
  between points \code{i} and \code{j}. Here \code{k} is the
  Epanechnikov kernel,
  \deqn{
    k(t) = \frac 3 { 4\delta} \max(0, 1 - \frac{t^2}{\delta^2}).
  }{
    k(t) = (3/(4*delta)) * max(0, 1 - t^2/delta^2).
  }
  Edge correction is performed using the border method
  (for the sake of computational efficiency):
  the estimate \eqn{g_i(r)}{g[i](r)} is set to \code{NA} if
  \eqn{r > b_i}{r > b[i]}, where \eqn{b_i}{b[i]}
  is the distance from point \eqn{i} to the boundary of the
  observation window.

  The smoothing bandwidth \eqn{\delta}{delta} may be specified.
  If not, it is chosen by Stoyan's rule of thumb
  \eqn{\delta = c/\hat\lambda}{delta = c/lambda}
  where \eqn{\hat\lambda = n/a}{lambda = n/a} is the estimated intensity
  and \eqn{c} is a constant, usually taken to be 0.15.
  The value of \eqn{c} is controlled by the argument \code{stoyan}.
}
\value{
  An object of class \code{"fv"}, see \code{\link{fv.object}},
  which can be plotted directly using \code{\link{plot.fv}}.
  Essentially a data frame containing columns
  \item{r}{the vector of values of the argument \eqn{r} 
    at which the function \eqn{K} has been  estimated
  }
  \item{theo}{the theoretical value \eqn{K(r) = \pi r^2}{K(r) = pi * r^2}
    or \eqn{L(r)=r} for a stationary Poisson process
  }
  together with columns containing the values of the
  local pair correlation function for each point in the pattern.
  Column \code{i} corresponds to the \code{i}th point.
  The last two columns contain the \code{r} and \code{theo} values.
}
\seealso{
  \code{\link{localK}},
  \code{\link{pcf}}
}
\examples{
  data(ponderosa)
  X <- ponderosa

  g <- localpcf(X)
  plot(g, main="local pair correlation functions for ponderosa", legend=FALSE)

  # plot only the local pair correlation function for point number 7
  plot(g, est007 ~ r)
}
\author{Adrian Baddeley
  \email{adrian@maths.uwa.edu.au}
  \url{http://www.maths.uwa.edu.au/~adrian/}
  and Rolf Turner
  \email{r.turner@auckland.ac.nz}
}
\keyword{spatial}
\keyword{nonparametric}