Estimates the pair correlation function of a point pattern using kernel methods.
# S3 method for ppp
pcf(X, ..., r = NULL,
adaptive=FALSE,
kernel="epanechnikov", bw=NULL, h=NULL,
bw.args=list(), stoyan=0.15, adjust=1,
correction=c("translate", "Ripley"),
divisor = c("r", "d", "a", "t"),
zerocor=c("weighted", "reflection", "convolution",
"bdrykern", "JonesFoster", "none"),
gref = NULL,
tau = 0,
fast = TRUE,
var.approx = FALSE,
domain=NULL,
ratio=FALSE, close=NULL)A function value table
(object of class "fv").
Essentially a data frame containing the variables
the vector of values of the argument \(r\) at which the pair correlation function \(g(r)\) has been estimated
vector of values equal to 1, the theoretical value of \(g(r)\) for the Poisson process
vector of values of \(g(r)\) estimated by translation correction
vector of values of \(g(r)\) estimated by Ripley isotropic correction
vector of approximate values of the variance of the estimate of \(g(r)\)
as required.
If ratio=TRUE then the return value also has two
attributes called "numerator" and "denominator"
which are "fv" objects
containing the numerators and denominators of each
estimate of \(g(r)\).
The return value also has an attribute "bw" giving the
smoothing bandwidth that was used, and an attribute "info"
containing details of the algorithm parameters.
The pair correlation function \(g(r)\)
is a summary of the dependence between points in a spatial point
process. The best intuitive interpretation is the following: the probability
\(p(r)\) of finding two points at locations \(x\) and \(y\)
separated by a distance \(r\) is equal to
$$
p(r) = \lambda^2 g(r) \,{\rm d}x \, {\rm d}y
$$
where \(\lambda\) is the intensity of the point process.
For a completely random (uniform Poisson) process,
\(p(r) = \lambda^2 \,{\rm d}x \, {\rm d}y\)
so \(g(r) = 1\).
Formally, the pair correlation function of a stationary point process
is defined by
$$
g(r) = \frac{K'(r)}{2\pi r}
$$
where \(K'(r)\) is the derivative of \(K(r)\), the
reduced second moment function (aka ``Ripley's \(K\) function'')
of the point process. See Kest for information
about \(K(r)\).
For a stationary Poisson process, the pair correlation function is identically equal to 1. Values \(g(r) < 1\) suggest inhibition between points; values greater than 1 suggest clustering.
This routine computes an estimate of \(g(r)\) by kernel smoothing.
If divisor="r" (the default), then the standard
kernel estimator (Stoyan and Stoyan, 1994, pages 284--285)
is used. By default, the recommendations of Stoyan and Stoyan (1994)
are followed exactly.
If divisor="d" then a modified estimator is used
(Guan, 2007): the contribution from
an interpoint distance \(d_{ij}\) to the
estimate of \(g(r)\) is divided by \(d_{ij}\)
instead of dividing by \(r\). This usually improves the
bias of the estimator when \(r\) is close to zero.
If divisor="a" then improved method of
Baddeley, Davies and Hazelton (2025) is used. The distances \(d_{ij}\)
are first converted to disc areas
\(a_{ij}=\pi d_{ij}^2\),
and smoothing is performed on the area scale,
then the result is back-transformed to the original scale.
If divisor="t" then the distances \(d_{ij}\)
are transformed to uniformity
using the reference pair correlation function gref
as described in Baddeley, Davies and Hazelton (2025).
If divisor is a function in the R language, then
it will be applied to the point pattern X and should return
one of the strings "r", "d", "a" or "t"
listed above. This option makes it possible to specify a rule
which decides which estimator to use, based on the data.
There is also a choice of spatial edge corrections (which are needed to avoid bias due to edge effects associated with the boundary of the spatial window):
If correction="translate" or correction="translation"
then the translation correction
is used. For divisor="r" the translation-corrected estimate
is given in equation (15.15), page 284 of Stoyan and Stoyan (1994).
If correction="Ripley" or correction="isotropic"
then Ripley's isotropic edge correction
is used. For divisor="r" the isotropic-corrected estimate
is given in equation (15.18), page 285 of Stoyan and Stoyan (1994).
If correction="none" then no edge correction is used,
that is, an uncorrected estimate is computed.
Multiple corrections can be selected. The default is
correction=c("translate", "Ripley").
Alternatively correction="all" selects all options;
correction="best" selects the option which has the best
statistical performance; correction="good" selects the option
which is the best compromise between statistical performance and speed
of computation.
Argument zerocor determines the correction
to the one-dimensional kernel-smoothed estimate
on the real number line, to correct bias close to the boundary \(r=0\).
The argument zerocor is passed to
densityBC.
Options include:
zerocor="none": no correction.
zerocor="convolution": the convolution, uniform or
renormalization kernel.
zerocor="weighted": the cut-and-normalization method.
zerocor="reflection":
the reflection method.
zerocor="bdrykern": the linear boundary kernel.
zerocor="JonesFoster":
the Jones-Foster modification of the linear boundary kernel.
The choice of smoothing kernel is controlled by the
argument kernel which is passed
to density.default.
The default is the Epanechnikov kernel, recommended by
Stoyan and Stoyan (1994, page 285).
The bandwidth of the smoothing kernel can be controlled by the
argument bw. Bandwidth is defined as the standard deviation
of the kernel; see the documentation for density.default.
For the Epanechnikov kernel with half-width h,
the argument bw is equivalent to \(h/\sqrt{5}\).
Stoyan and Stoyan (1994, page 285) recommend using the Epanechnikov
kernel with support \([-h,h]\) chosen by the rule of thumn
\(h = c/\sqrt{\lambda}\),
where \(\lambda\) is the (estimated) intensity of the
point process, and \(c\) is a constant in the range from 0.1 to 0.2.
See equation (15.16).
If bw is missing or NULL,
then this rule of thumb will be applied.
The argument stoyan determines the value of \(c\).
The smoothing bandwidth that was used in the calculation is returned
as an attribute of the final result.
The argument bw can be
missing or null. In this case, the default value for bw
is "stoyan" when adaptive=FALSE
and "bw.abram" when adaptive=TRUE.
a single numeric value giving the bandwidth.
a character string specifying a bandwidth selection rule.
String names of rules applicable when adaptive=FALSE
include "stoyan", "fiksel"
and any rules recognised by density.default.
String names applicable when adaptive=TRUE include
"bw.abram" and "bw.pow".
a function that computes the selected bandwidth.
If adaptive=FALSE, the function bw will be applied to the
point pattern X to determine the bandwidth. Examples include
bw.pcf and bw.stoyan.
The function bw should accept the point pattern X
as its first argument. Additional arguments to bw may be
specified in the list bw.args. If bw recognises any
of the arguments
kernel, correction, divisor, zerocor
and adaptive, then these arguments will be passed to
bw as well.
The function bw should return a single
numeric value.
If adaptive=TRUE, the function bw will be applied to
the vector of pairwise distances between data points (or the
transformed distances if divisor="a" or divisor="t").
Examples include bw.abram.default
and bw.pow.
The function bw should accept the vector of pairwise
distances as its first argument. Additional arguments to bw may be
specified in the list bw.args.
Note that if bw.args is a function, it will be applied to
the point pattern X to determine the list of arguments
(whether adaptive is TRUE or FALSE).
The argument r is the vector of values for the
distance \(r\) at which \(g(r)\) should be evaluated.
There is a sensible default.
If it is specified, r must be a vector of increasing numbers
starting from r[1] = 0,
and max(r) must not exceed half the diameter of
the window.
If the argument domain is given, estimation will be restricted
to this region. That is, the estimate of
\(g(r)\) will be based on pairs of points in which the first point lies
inside domain and the second point is unrestricted.
The argument domain
should be a window (object of class "owin") or something acceptable to
as.owin. It must be a subset of the
window of the point pattern X.
To compute a confidence band for the true value of the
pair correlation function, use lohboot.
If var.approx = TRUE, the variance of the
estimate of the pair correlation will also be calculated using
an analytic approximation (Illian et al, 2008, page 234)
which is valid for stationary point processes which are not
too clustered. This calculation is not yet implemented when
the argument domain is given.
If fast=TRUE, the calculation uses the Fast Fourier Transform
to the maximum extent possible for the chosen boundary correction.
If fast=FALSE (the default), the entire calculation uses
analytic formulas written in C code.
Baddeley, A., Davies, T.M. and Hazelton, M.L. (2025) An improved estimator of the pair correlation function of a spatial point process. Biometrika, to appear.
Guan, Y. (2007) A least-squares cross-validation bandwidth selection approach in pair correlation function estimation. Statistics and Probability Letters 77 (18) 1722--1729.
Illian, J., Penttinen, A., Stoyan, H. and Stoyan, D. (2008) Statistical Analysis and Modelling of Spatial Point Patterns. Wiley.
Stoyan, D. and Stoyan, H. (1994) Fractals, random shapes and point fields: methods of geometrical statistics. John Wiley and Sons.
densityBC,
Kest,
pcf,
density.default,
bw.stoyan,
bw.pcf,
lohboot.
pr <- pcf(redwood, divisor="r")
plot(pr, main="pair correlation function for redwoods")
# compare estimates
pd <- pcf(redwood, divisor="d")
pa <- pcf(redwood, divisor="a")
plot(pr, cbind(iso, theo) ~ r, col=c("red", "black"),
ylim.covers=0, main="Estimates of PCF",
lwd=c(2,1), lty=c(1,3), legend=FALSE)
plot(pd, iso ~ r, col="blue", lwd=2, add=TRUE)
plot(pa, iso ~ r, col="green", lwd=2, add=TRUE)
legend("center", col=c("red", "blue", "green"), lty=1, lwd=2,
legend=c("divisor=r","divisor=d", "divisor=a"))
Run the code above in your browser using DataLab