Simulate Point Process Models using the Metropolis-Hastings Algorithm.
Generates a random point pattern, simulated from a chosen point process model, using the Metropolis-Hastings algorithm.
## Default S3 method: rmh(model, start=NULL, control=default.rmhcontrol(model), ..., nsim=1, drop=TRUE, saveinfo=TRUE, verbose=TRUE, snoop=FALSE)
model |
Data specifying the point process model that is to be simulated. |
start |
Data determining the initial state of the algorithm. |
control |
Data controlling the iterative behaviour and termination of the algorithm. |
... |
Further arguments passed to |
nsim |
Number of simulated point patterns that should be generated. |
drop |
Logical. If |
saveinfo |
Logical value indicating whether to save auxiliary information. |
verbose |
Logical value indicating whether to print progress reports. |
snoop |
Logical. If |
This function generates simulated realisations from any of a range of
spatial point processes, using the Metropolis-Hastings algorithm.
It is the default method for the generic function rmh.
This function executes a Metropolis-Hastings algorithm with birth, death and shift proposals as described in Geyer and Moller (1994).
The argument model specifies the point process model to be
simulated. It is either a list, or an object of class
"rmhmodel", with the following components:
A character string specifying the choice of interpoint interaction for the point process.
Parameter values for the conditional intensity function.
(Optional) window in which the pattern is
to be generated. An object of class "owin",
or data acceptable to as.owin.
Data specifying the spatial trend in the model, if it has a trend.
This may be a function, a pixel image (of class "im"),
(or a list of functions or images if the model
is multitype).
If the trend is a function or functions,
any auxiliary arguments ... to rmh.default
will be passed to these functions, which
should be of the form function(x, y, ...).
List of possible types, for a multitype point process.
For full details of these parameters, see rmhmodel.default.
The argument start determines the initial state of the
Metropolis-Hastings algorithm. It is either NULL,
or an object of class "rmhstart",
or a list with the following components:
Number of points in the initial point pattern.
A single integer, or a vector of integers giving the
numbers of points of each type in a multitype point pattern.
Incompatible with x.start.
Initial point pattern configuration.
Incompatible with n.start.
x.start may be a point pattern (an
object of class "ppp"), or data which can be coerced
to this class by as.ppp, or an object with
components x and y, or a two-column matrix.
In the last two cases, the window for the pattern is determined
by model$w.
In the first two cases, if model$w is also present,
then the final simulated pattern will be clipped to
the window model$w.
For full details of these parameters, see rmhstart.
The third argument control controls the simulation
procedure (including conditional simulation),
iterative behaviour, and termination of the
Metropolis-Hastings algorithm. It is either NULL, or
a list, or an object of class "rmhcontrol", with components:
The probability of proposing a “shift” (as opposed to a birth or death) in the Metropolis-Hastings algorithm.
The conditional probability of proposing a death (rather than a birth) given that birth/death has been chosen over shift.
The number of repetitions or iterations to be made by the Metropolis-Hastings algorithm. It should be large.
Either a numerical expansion factor, or
a window (object of class "owin"). Indicates that
the process is to be simulated on a larger domain than the
original data window w, then clipped to w
when the algorithm has finished.
The default is to expand the simulation window if the model is stationary and non-Poisson (i.e. it has no trend and the interaction is not Poisson) and not to expand in all other cases.
If the model has a trend, then in order for expansion to be feasible, the trend must be given either as a function, or an image whose bounding box is large enough to contain the expanded window.
A logical scalar; if periodic is TRUE
we simulate a process on the torus formed by identifying
opposite edges of a rectangular window.
A vector of probabilities (summing to 1) to be used in assigning a random type to a new point.
A logical scalar specifying whether to condition on the number of points of each type.
An integer specifying how often “progress reports” (which consist simply of the number of repetitions completed) should be printed out. If nverb is left at 0, the default, the simulation proceeds silently.
If this argument is present, then
conditional simulation will be performed, and x.cond
specifies the conditioning points and the type of conditioning.
If these values are specified, then
intermediate states of the simulation algorithm will be saved
every nsave iterations, after an initial burn-in period of
nburn iterations.
Logical flag indicating whether to save the transition history of the simulations.
For full details of these parameters, see rmhcontrol.
The control parameters can also be given in the ... arguments.
A point pattern (an object of class "ppp", see
ppp.object) or a list of point patterns.
The returned value has an attribute info containing
modified versions of the arguments
model, start, and control which together specify
the exact simulation procedure. The info attribute can be
printed (and is printed automatically by summary.ppp).
For computational efficiency, the info attribute can be omitted
by setting saveinfo=FALSE.
The value of .Random.seed at the start
of the simulations is also saved and returned as an attribute
seed.
If the argument track=TRUE was given (see rmhcontrol),
the transition history of the algorithm
is saved, and returned as an attribute history. The transition
history is a data frame containing a factor proposaltype
identifying the proposal type (Birth, Death or Shift) and
a logical vector accepted indicating whether the proposal was
accepted.
The data frame also has columns numerator, denominator
which give the numerator and denominator of the Hastings ratio for
the proposal.
If the argument nsave was given (see rmhcontrol),
the return value has an attribute saved which is a list of
point patterns, containing the intermediate states of the algorithm.
There are several kinds of conditional simulation.
Simulation conditional upon the number of points,
that is, holding the number of points fixed.
To do this, set control$p (the probability of a shift) equal to 1.
The number of points is then determined by the starting state, which
may be specified either by setting start$n.start to be a
scalar, or by setting the initial pattern start$x.start.
In the case of multitype processes, it is possible to simulate the
model conditionally upon the number of points of each type,
i.e. holding the number of points of each type
to be fixed. To do this, set control$p equal to 1
and control$fixall to be TRUE.
The number of points is then determined by the starting state, which
may be specified either by setting start$n.start to be an
integer vector, or by setting the initial pattern start$x.start.
Simulation
conditional on the configuration observed in a sub-window,
that is, requiring that, inside a specified sub-window V,
the simulated pattern should agree with a specified point pattern
y.To do this, set control$x.cond to equal the
specified point pattern y, making sure that it is an object of class
"ppp" and that the window Window(control$x.cond)
is the conditioning window V.
Simulation conditional on the presence of specified points,
that is, requiring that the simulated pattern should include a
specified set of points. This is simulation from the Palm
distribution of the point process given a pattern y.
To do this, set control$x.cond to be a
data.frame containing the coordinates (and marks,
if appropriate) of the specified points.
For further information, see rmhcontrol.
Note that, when we simulate conditionally on the number of points, or conditionally on the number of points of each type, no expansion of the window is possible.
If snoop = TRUE, an interactive debugger is activated.
On the current plot device, the debugger displays the current
state of the Metropolis-Hastings algorithm together with
the proposed transition to the next state.
Clicking on this graphical display (using the left mouse button)
will re-centre the display at the clicked location.
Surrounding this graphical display is an array of boxes representing
different actions.
Clicking on one of the action boxes (using the left mouse button)
will cause the action to be performed.
Debugger actions include:
Zooming in or out
Panning (shifting the field of view) left, right, up or down
Jumping to the next iteration
Skipping 10, 100, 1000, 10000 or 100000 iterations
Jumping to the next Birth proposal (etc)
Changing the fate of the proposal (i.e. changing whether the proposal is accepted or rejected)
Dumping the current state and proposal to a file
Printing detailed information at the terminal
Exiting the debugger (so that the simulation algorithm continues without further interruption).
Right-clicking the mouse will also cause the debugger to exit.
There is never a guarantee that the Metropolis-Hastings algorithm has converged to its limiting distribution.
If start$x.start is specified then expand is set equal to 1
and simulation takes place in Window(x.start). Any specified
value for expand is simply ignored.
The presence of both a component w of model and a
non-null value for Window(x.start) makes sense ONLY if w
is contained in Window(x.start).
For multitype processes make sure that, even if there is to be no trend corresponding to a particular type, there is still a component (a NULL component) for that type, in the list.
In theory, any finite point process model can be simulated using the Metropolis-Hastings algorithm, provided the conditional intensity is uniformly bounded.
In practice, the list of point process models that can be simulated using
rmh.default is limited to those that have been implemented
in the package's internal C code. More options will be added in the future.
Note that the lookup conditional intensity function
permits the simulation (in theory, to any desired degree
of approximation) of any pairwise interaction process for
which the interaction depends only on the distance between
the pair of points.
If the user wants the simulation to be exactly reproducible
(e.g. for a figure in a journal article, where it is useful to
have the figure consistent from draft to draft) then the state of
the random number generator should be set before calling
rmh.default. This can be done either by calling
set.seed or by assigning a value to
.Random.seed. In the examples below, we use
set.seed.
If a simulation has been performed and the user now wants to
repeat it exactly, the random seed should be extracted from
the simulated point pattern X by seed <- attr(x, "seed"),
then assigned to the system random nunber state by
.Random.seed <- seed before calling rmh.default.
Adrian Baddeley Adrian.Baddeley@curtin.edu.au
and Rolf Turner r.turner@auckland.ac.nz
Baddeley, A. and Turner, R. (2000) Practical maximum pseudolikelihood for spatial point patterns. Australian and New Zealand Journal of Statistics 42, 283 – 322.
Diggle, P. J. (2003) Statistical Analysis of Spatial Point Patterns (2nd ed.) Arnold, London.
Diggle, P.J. and Gratton, R.J. (1984) Monte Carlo methods of inference for implicit statistical models. Journal of the Royal Statistical Society, series B 46, 193 – 212.
Diggle, P.J., Gates, D.J., and Stibbard, A. (1987) A nonparametric estimator for pairwise-interaction point processes. Biometrika 74, 763 – 770.
Geyer, C.J. and Moller, J. (1994) Simulation procedures and likelihood inference for spatial point processes. Scandinavian Journal of Statistics 21, 359–373.
Geyer, C.J. (1999) Likelihood Inference for Spatial Point Processes. Chapter 3 in O.E. Barndorff-Nielsen, W.S. Kendall and M.N.M. Van Lieshout (eds) Stochastic Geometry: Likelihood and Computation, Chapman and Hall / CRC, Monographs on Statistics and Applied Probability, number 80. Pages 79–140.
if(interactive()) {
nr <- 1e5
nv <- 5000
ns <- 200
} else {
nr <- 20
nv <- 5
ns <- 20
oldopt <- spatstat.options()
spatstat.options(expand=1.05)
}
set.seed(961018)
# Strauss process.
mod01 <- list(cif="strauss",par=list(beta=2,gamma=0.2,r=0.7),
w=c(0,10,0,10))
X1.strauss <- rmh(model=mod01,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X1.strauss)
# Strauss process, conditioning on n = 42:
X2.strauss <- rmh(model=mod01,start=list(n.start=42),
control=list(p=1,nrep=nr,nverb=nv))
# Tracking algorithm progress:
# (a) saving intermediate states:
X <- rmh(model=mod01,start=list(n.start=ns),
control=list(nrep=nr, nsave=nr/5, nburn=nr/2))
Saved <- attr(X, "saved")
plot(Saved)
# (b) inspecting transition history:
X <- rmh(model=mod01,start=list(n.start=ns),
control=list(nrep=nr, track=TRUE))
History <- attr(X, "history")
head(History)
# Hard core process:
mod02 <- list(cif="hardcore",par=list(beta=2,hc=0.7),w=c(0,10,0,10))
X3.hardcore <- rmh(model=mod02,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X3.hardcore)
# Strauss process equal to pure hardcore:
mod02s <- list(cif="strauss",par=list(beta=2,gamma=0,r=0.7),w=c(0,10,0,10))
X3.strauss <- rmh(model=mod02s,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Strauss process in a polygonal window.
x <- c(0.55,0.68,0.75,0.58,0.39,0.37,0.19,0.26,0.42)
y <- c(0.20,0.27,0.68,0.99,0.80,0.61,0.45,0.28,0.33)
mod03 <- list(cif="strauss",par=list(beta=2000,gamma=0.6,r=0.07),
w=owin(poly=list(x=x,y=y)))
X4.strauss <- rmh(model=mod03,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X4.strauss)
# Strauss process in a polygonal window, conditioning on n = 80.
X5.strauss <- rmh(model=mod03,start=list(n.start=ns),
control=list(p=1,nrep=nr,nverb=nv))
# Strauss process, starting off from X4.strauss, but with the
# polygonal window replace by a rectangular one. At the end,
# the generated pattern is clipped to the original polygonal window.
xxx <- X4.strauss
Window(xxx) <- as.owin(c(0,1,0,1))
X6.strauss <- rmh(model=mod03,start=list(x.start=xxx),
control=list(nrep=nr,nverb=nv))
# Strauss with hardcore:
mod04 <- list(cif="straush",par=list(beta=2,gamma=0.2,r=0.7,hc=0.3),
w=c(0,10,0,10))
X1.straush <- rmh(model=mod04,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Another Strauss with hardcore (with a perhaps surprising result):
mod05 <- list(cif="straush",par=list(beta=80,gamma=0.36,r=45,hc=2.5),
w=c(0,250,0,250))
X2.straush <- rmh(model=mod05,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Pure hardcore (identical to X3.strauss).
mod06 <- list(cif="straush",par=list(beta=2,gamma=1,r=1,hc=0.7),
w=c(0,10,0,10))
X3.straush <- rmh(model=mod06,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Soft core:
w <- c(0,10,0,10)
mod07 <- list(cif="sftcr",par=list(beta=0.8,sigma=0.1,kappa=0.5),
w=c(0,10,0,10))
X.sftcr <- rmh(model=mod07,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.sftcr)
# Area-interaction process:
mod42 <- rmhmodel(cif="areaint",par=list(beta=2,eta=1.6,r=0.7),
w=c(0,10,0,10))
X.area <- rmh(model=mod42,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.area)
# Triplets process
modtrip <- list(cif="triplets",par=list(beta=2,gamma=0.2,r=0.7),
w=c(0,10,0,10))
X.triplets <- rmh(model=modtrip,
start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.triplets)
# Multitype Strauss:
beta <- c(0.027,0.008)
gmma <- matrix(c(0.43,0.98,0.98,0.36),2,2)
r <- matrix(c(45,45,45,45),2,2)
mod08 <- list(cif="straussm",par=list(beta=beta,gamma=gmma,radii=r),
w=c(0,250,0,250))
X1.straussm <- rmh(model=mod08,start=list(n.start=ns),
control=list(ptypes=c(0.75,0.25),nrep=nr,nverb=nv))
if(interactive()) plot(X1.straussm)
# Multitype Strauss conditioning upon the total number
# of points being 80:
X2.straussm <- rmh(model=mod08,start=list(n.start=ns),
control=list(p=1,ptypes=c(0.75,0.25),nrep=nr,
nverb=nv))
# Conditioning upon the number of points of type 1 being 60
# and the number of points of type 2 being 20:
X3.straussm <- rmh(model=mod08,start=list(n.start=c(60,20)),
control=list(fixall=TRUE,p=1,ptypes=c(0.75,0.25),
nrep=nr,nverb=nv))
# Multitype Strauss hardcore:
rhc <- matrix(c(9.1,5.0,5.0,2.5),2,2)
mod09 <- list(cif="straushm",par=list(beta=beta,gamma=gmma,
iradii=r,hradii=rhc),w=c(0,250,0,250))
X.straushm <- rmh(model=mod09,start=list(n.start=ns),
control=list(ptypes=c(0.75,0.25),nrep=nr,nverb=nv))
# Multitype Strauss hardcore with trends for each type:
beta <- c(0.27,0.08)
tr3 <- function(x,y){x <- x/250; y <- y/250;
exp((6*x + 5*y - 18*x^2 + 12*x*y - 9*y^2)/6)
}
# log quadratic trend
tr4 <- function(x,y){x <- x/250; y <- y/250;
exp(-0.6*x+0.5*y)}
# log linear trend
mod10 <- list(cif="straushm",par=list(beta=beta,gamma=gmma,
iradii=r,hradii=rhc),w=c(0,250,0,250),
trend=list(tr3,tr4))
X1.straushm.trend <- rmh(model=mod10,start=list(n.start=ns),
control=list(ptypes=c(0.75,0.25),
nrep=nr,nverb=nv))
if(interactive()) plot(X1.straushm.trend)
# Multitype Strauss hardcore with trends for each type, given as images:
bigwin <- square(250)
i1 <- as.im(tr3, bigwin)
i2 <- as.im(tr4, bigwin)
mod11 <- list(cif="straushm",par=list(beta=beta,gamma=gmma,
iradii=r,hradii=rhc),w=bigwin,
trend=list(i1,i2))
X2.straushm.trend <- rmh(model=mod11,start=list(n.start=ns),
control=list(ptypes=c(0.75,0.25),expand=1,
nrep=nr,nverb=nv))
# Diggle, Gates, and Stibbard:
mod12 <- list(cif="dgs",par=list(beta=3600,rho=0.08),w=c(0,1,0,1))
X.dgs <- rmh(model=mod12,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.dgs)
# Diggle-Gratton:
mod13 <- list(cif="diggra",
par=list(beta=1800,kappa=3,delta=0.02,rho=0.04),
w=square(1))
X.diggra <- rmh(model=mod13,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.diggra)
# Fiksel:
modFik <- list(cif="fiksel",
par=list(beta=180,r=0.15,hc=0.07,kappa=2,a= -1.0),
w=square(1))
X.fiksel <- rmh(model=modFik,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.fiksel)
# Geyer:
mod14 <- list(cif="geyer",par=list(beta=1.25,gamma=1.6,r=0.2,sat=4.5),
w=c(0,10,0,10))
X1.geyer <- rmh(model=mod14,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X1.geyer)
# Geyer; same as a Strauss process with parameters
# (beta=2.25,gamma=0.16,r=0.7):
mod15 <- list(cif="geyer",par=list(beta=2.25,gamma=0.4,r=0.7,sat=10000),
w=c(0,10,0,10))
X2.geyer <- rmh(model=mod15,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
mod16 <- list(cif="geyer",par=list(beta=8.1,gamma=2.2,r=0.08,sat=3))
data(redwood)
X3.geyer <- rmh(model=mod16,start=list(x.start=redwood),
control=list(periodic=TRUE,nrep=nr,nverb=nv))
# Geyer, starting from the redwood data set, simulating
# on a torus, and conditioning on n:
X4.geyer <- rmh(model=mod16,start=list(x.start=redwood),
control=list(p=1,periodic=TRUE,nrep=nr,nverb=nv))
# Lookup (interaction function h_2 from page 76, Diggle (2003)):
r <- seq(from=0,to=0.2,length=101)[-1] # Drop 0.
h <- 20*(r-0.05)
h[r<0.05] <- 0
h[r>0.10] <- 1
mod17 <- list(cif="lookup",par=list(beta=4000,h=h,r=r),w=c(0,1,0,1))
X.lookup <- rmh(model=mod17,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
if(interactive()) plot(X.lookup)
# Strauss with trend
tr <- function(x,y){x <- x/250; y <- y/250;
exp((6*x + 5*y - 18*x^2 + 12*x*y - 9*y^2)/6)
}
beta <- 0.3
gmma <- 0.5
r <- 45
modStr <- list(cif="strauss",par=list(beta=beta,gamma=gmma,r=r),
w=square(250), trend=tr)
X1.strauss.trend <- rmh(model=modStr,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Baddeley-Geyer
r <- seq(0,0.2,length=8)[-1]
gmma <- c(0.5,0.6,0.7,0.8,0.7,0.6,0.5)
mod18 <- list(cif="badgey",par=list(beta=4000, gamma=gmma,r=r,sat=5),
w=square(1))
X1.badgey <- rmh(model=mod18,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
mod19 <- list(cif="badgey",
par=list(beta=4000, gamma=gmma,r=r,sat=1e4),
w=square(1))
set.seed(1329)
X2.badgey <- rmh(model=mod18,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Check:
h <- ((prod(gmma)/cumprod(c(1,gmma)))[-8])^2
hs <- stepfun(r,c(h,1))
mod20 <- list(cif="lookup",par=list(beta=4000,h=hs),w=square(1))
set.seed(1329)
X.check <- rmh(model=mod20,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# X2.badgey and X.check will be identical.
mod21 <- list(cif="badgey",par=list(beta=300,gamma=c(1,0.4,1),
r=c(0.035,0.07,0.14),sat=5), w=square(1))
X3.badgey <- rmh(model=mod21,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Same result as Geyer model with beta=300, gamma=0.4, r=0.07,
# sat = 5 (if seeds and control parameters are the same)
# Or more simply:
mod22 <- list(cif="badgey",
par=list(beta=300,gamma=0.4,r=0.07, sat=5),
w=square(1))
X4.badgey <- rmh(model=mod22,start=list(n.start=ns),
control=list(nrep=nr,nverb=nv))
# Same again --- i.e. the BadGey model includes the Geyer model.
# Illustrating scalability.
if(FALSE) {
M1 <- rmhmodel(cif="strauss",par=list(beta=60,gamma=0.5,r=0.04),w=owin())
set.seed(496)
X1 <- rmh(model=M1,start=list(n.start=300))
M2 <- rmhmodel(cif="strauss",par=list(beta=0.6,gamma=0.5,r=0.4),
w=owin(c(0,10),c(0,10)))
set.seed(496)
X2 <- rmh(model=M2,start=list(n.start=300))
chk <- affine(X1,mat=diag(c(10,10)))
all.equal(chk,X2,check.attributes=FALSE)
# Under the default spatstat options the foregoing all.equal()
# will yield TRUE. Setting spatstat.options(scalable=FALSE) and
# re-running the code will reveal differences between X1 and X2.
}
if(!interactive()) spatstat.options(oldopt)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.