RELEASE NOTES
spatstat 1.40-0
31 December 2014
We thank Markus Herrmann, Peter Kovesi, Andrew Lister, Enrique Miranda, Tuomas Rajala, Brian Ripley, Dominic Schuhmacher and Maxime Woringer for contributions.
OVERVIEW
-
Important bug fixes.
-
Mathematical operators now apply to images.
-
Parametric estimates of relative risk from fitted point process models.
-
Standard errors for relative risk (parametric and non-parametric).
-
Kernel smoothing and rose diagrams for angular data.
-
Perceptually uniform colour maps.
-
Hierarchical interactions for multitype patterns.
-
Hard core parameters in all interactions no longer need to be specified
and will be estimated from data. -
Improvements to analysis of deviance and model selection.
-
New datasets.
-
New vignette, summarising all datasets installed with spatstat.
-
Tests and diagnostics now include a Monte Carlo option.
-
Faster checking of large datasets.
-
Faster simulations.
-
Code for drawing diagrams (arrows, scale bars).
-
Version nickname: ‘Do The Maths’
NEW DATASETS
-
bdspots
Breakdown spots on microelectronic capacitor electrodes.
Generously contributed by Prof Enrique Miranda. -
Kovesi
Colour maps with perceptually uniform contrast.
Generously contributed by Peter Kovesi.
NEW FUNCTIONS
-
Mathematical operations are now supported for images.
For example: alpha <- atan(bei.extra$grad) * 180/pi
See help(Math.im) -
relrisk.ppm
Spatially-varying probabilities of different types of points
predicted by a fitted point process model. -
circdensity
Kernel density estimate for angular data -
rose
Rose diagram (rose of directions) for angular data -
nnorient
Nearest neighbour orientation distribution. -
AIC.ppm
Calculate AIC of a Gibbs model using Takeuchi’s rule. -
interp.colours
Interpolate a sequence of colour values. -
anyDuplicated.ppp, anyDuplicated.ppx
Fast replacements for any(duplicated(x)) for point patterns. -
textstring, onearrow, yardstick
Objects representing a text string, an arrow, or a scale bar,
for use in drawing spatial diagrams. -
plot.imlist, image.imlist, contour.imlist
Methods for the new class ‘imlist’ -
[<-.layered, [[<-.layered
More support for class ‘layered’
SIGNIFICANT USER-VISIBLE CHANGES
-
(vignettes)
New vignette ‘datasets’ summarises all the datasets installed
with the spatstat package. -
relrisk
The function relrisk is now generic, with methods for ppp and ppm.
New argument ‘relative’ specifies whether to calculate the relative risk
or the absolute probability of each type of point.
New argument ‘se’ specifies whether to calculate standard errors. -
plot.im
The default colour map for plotting images,
specified by spatstat.options(‘image.colfun’),
has been changed to a perceptually uniform map. -
DiggleGratton, Fiksel, MultiHard, MultiStraussHard
The hard core distance parameters in these models
can now be omitted by the user, and will be estimated automatically
from data (by the ‘self-starting’ feature of interactions).
This was already true of Hardcore and StraussHard. -
Hybrid
Hybrid models now apply the ‘self-starting’ feature
to each component model. -
anova.ppm
Can now reconcile models fitted using different dummy points,
different values of ‘rbord’, different values of ‘use.gam’, etc. -
profilepl
New argument ‘aic’ makes it possible to optimise the parameters
by minimising AIC. -
profilepl
No longer requires values for parameters which are ‘optional’
(such as the hard core distance). -
rmh, simulate.ppm, rmh.ppm, rmh.default
The Metropolis-Hastings algorithm now starts by deleting any points
in the initial state that are ‘illegal’ (i.e. whose conditional intensity
is equal to zero). This ensures that the result of rmh
never contains illegal points. -
runifpoint, rpoispp, rStrauss, rHardcore, rStraussHard,
rDiggleGratton, rDGS, runifdisc, rpoint, rMaternI, rMaternII, rSSI,
rPoissonCluster, rGaussPoisson, rstrat, rsyst, rcell, rthin, rjitter,
rNeymanScott, rMatClust, rThomas, rCauchy, rVarGamma, rmpoint, rmpoispp,
runifpointOnLines, rpoisppOnLines, runiflpp, rpoislpp, runifpointx,
rpoisppx, runifpoint3, rpoispp3
These random point pattern generators now have an argument ‘nsim’
specifying the number of simulated realisations to be generated. -
pairorient
New argument ‘cumulative’.
New algorithm to compute kernel estimate of probability density.
Default behaviour changed.
Argument ‘units’ has been renamed ‘unit’ for consistency.
Labels and descriptions of columns have been corrected. -
predict.ppm
New syntax (backward-compatible).
New argument ‘se’ replaces option ‘type=”se”’.
Old argument ‘total’ is deprecated: use ‘window’ and set ‘type=”count”’. -
cdf.test
The methods for class ‘ppm’ and ‘lppm’ now handle Gibbs models
and perform a Monte Carlo test in this case. -
lurking, diagnose.ppm
Lurking variable plot can now include simulation envelopes. -
rmh.ppm
New argument ‘w’ determines the window in which the simulated pattern
is generated. -
ppp
Accelerated. -
Gcom, Gres
When conditional=TRUE and restrict=TRUE, the Hanisch estimate
was not calculated exactly as described in Appendix E.1 of
Baddeley, Rubak and Moller (2011). The intensity was estimated
on the full window rather than the eroded window.
Fixed. -
step, drop1, add1, extractAIC
The AIC of a Gibbs model is now calculated using Takeuchi’s rule
for the degrees of freedom. -
model.matrix.ppm, model.matrix.kppm
New argument ‘Q’ allows prediction at any desired locations. -
vcov.ppm
New argument ‘fine’ gives more control over computation. -
predict.ppm
For multitype models, when the result is a list of images,
the names of list entries are now identical to the mark levels
(e.g. “hickory” instead of “markhickory”) -
print.slrm
Output now respects options(‘width’) -
image.listof
New argument ‘ribmar’ controls margin space around the ribbon
when equal.ribbon=TRUE. -
integral.im
New argument ‘domain’ specifies the domain of integration. -
plot.fasp
New argument ‘transpose’ allows rows and columns to be exchanged. -
plot.im
The list ‘ribargs’ can now include the parameter ‘labels’. -
rmh, rpoint, rpoispp, rmpoint, rmpoispp
Accelerated, for inhomogeneous processes. -
stienen
Now recognises the parameter ‘lwd’. -
suffstat
Accelerated (also affects ppm with method=’ho’). -
Poisson, AreaInter, BadGey, Concom, DiggleGatesStibbard, DiggleGratton,
Fiksel, Geyer, Hardcore, Hybrid, LennardJones, MultiHard, MultiStrauss,
MultiStraussHard, OrdThresh, Ord, PairPiece, Pairwise, SatPiece,
Saturated, Softcore, Strauss, StraussHard, Triplets
These functions can now be printed (by typing the function name)
to give a sensible description of the required syntax. -
fitin
A plot of the fitted interpoint interaction of a point process model
e.g. plot(fitin(ppm(swedishpines ~ 1, Strauss(9))))
now shows the unit of length on the x-axis. -
fitin
Plots of the fitted interpoint interaction are now possible
for some higher-order interactions such as Geyer and AreaInter. -
anova.ppm
New argument ‘warn’ to suppress warnings. -
rmhmodel.ppm
Argument ‘win’ renamed ‘w’ for consistency with other functions. -
print.ppm
Printed output for the fitted regular parameters
now respects options(‘digits’). -
print.ppm, print.summary.ppm
Output now respects options(‘width’) and spatstat.options(‘terse’) -
print.ppm
By default, standard errors are not printed
for a model fitted with method=”logi” (due to computational load) -
plot.profilepl
Now recognises ‘lty’, ‘lwd’, ‘col’ etc -
vesicles, gorillas
Some of the raw data files for these datasets are also installed in spatstat
for demonstration and training purposes.
BUG FIXES
-
rmh, rmh.ppm, rmh.default, simulate.ppm
The result of simulating a model with a hard core
did not necessarily respect the hard core constraint,
and simulation of a model with strong inhibition
did not necessarily converge.
This only happened if the first order trend was large,
the starting state (n.start or x.start) was not given,
and the number of iterations (nrep) was not very large.
It occurred because of a poor choice for the default starting state.
Bug was present since about 2010.
Fixed. -
markcorrint
Results were completely incorrect.
Bug introduced in spatstat 1.39-0, october 2014.
Fixed. -
Kinhom
Ignored argument ‘reciplambda2’ in some cases.
Bug introduced in spatstat 1.39-0, october 2014.
Fixed. -
relrisk
When at=”pixels”, a small fraction of pixel values were sometimes
wildly inaccurate, due to numerical errors. This affected the
range of values in the result, and therefore the appearance of plots.
Fixed. -
model.images
Crashed if the model was multitype.
Fixed. -
profilepl
Crashed in some cases when the interaction was multitype.
[Spotted by Andrew Lister.]
Fixed. -
profilepl
Crashed if the model involved covariates that were not
given in a ‘data’ argument.
Fixed. -
envelope.ppm
Crashed if global=TRUE and savefuns=TRUE.
Fixed. -
setminus.owin
Crashed if the result was empty and the input was polygonal.
Fixed. -
predict.ppm
Crashed sometimes when type=”cif” and ngrid was very large.
Fixed. -
pixelquad
If X was a multitype point pattern, the result was mangled.
Fixed. -
relrisk
Did not accept a character string value for the argument ‘case’.
Fixed. -
intensity.ppm
Format of result was incorrect for ppm(Y ~ 1) where Y is multitype.
Fixed. -
$<-.hyperframe
Columns containing character values were converted to factors.
Fixed. -
clickppp
Sometimes filled the window in solid black colour..
Fixed. -
plot.psp
Ignored ‘show.all’ in some cases.
Fixed. -
plot.ppp
Warned about NA values amongst the marks, even if there were no NA’s
in the column(s) of marks selected by the argument ‘which.marks’.
Fixed. -
stienen
Did not suppress the border circles when border=FALSE.
Fixed.
Release notes are available in raw text format here.