CHANGES IN spatstat VERSION 2.2-0
OVERVIEW
o We thank Warick Brown, Achmad Choiruddin, Jean-Francois Coeurjolly,
Andrea Gilardi, Yongtao Guan, Abdollah Jalilian, Hank Stevens
and Rasmus Waagepetersen for contributions.
o Conditional simulation in kppm.
o Simulation of the product shot noise Cox process.
o Information criteria for model selection in kppm
o Estimation of the spatial covariance function of a pixel image
o Modified handling of covariates in slrm
o Buffer tessellation
o New function for jittering point patterns on a network.
o Extensions to 'rhohat'
o densityfun.ppp handles query points outside original window
o Extension to 'discretise'.
o Improvement to densityEqualSplit
o summary method for spatial logistic regression models
o New options for distmap.psp
o Improved output in summary.mppm
o Bug fix in nncross
o Bug fix in density.lpp
NEW FUNCTIONS
o bufftess
Distance buffer tessellation
o ic
Information criteria for model selection in ppm and kppm.
Kindly contributed by Achmad Choiruddin, Jean-Francois Coeurjolly
and Rasmus Waagepetersen.
o rPSNCP
Generate simulated realisations of the product shot noise Cox process.
Contributed by Abdollah Jalilian, Yongtao Guan and Rasmus Waagepetersen.
o spatcov
Estimate the spatial covariance function of a pixel image.
o summary.slrm, print.summary.slrm
Summary method for spatial logistic regression models
o coef.summary.slrm
Print the fitted coefficients, confidence interval and p-values
for a spatial logistic regression model.
o pairMean
Compute the mean of a specified function of interpoint distance
between random points in a window.
o rjitterlpp
Apply random displacements to the points on a linear network.
SIGNIFICANT USER-VISIBLE CHANGES
o simulate.kppm
Conditional simulation of the model, given a fixed number of points,
is now supported using the new arguments 'n.cond' and 'w.cond'.
o densityfun.ppp
The resulting function can now handle query points which lie
outside the window of the original data,
and has argument 'drop=TRUE' which specifies how to handle them.
o rpoint
New argument 'forcewin' forces the code to use the window 'win'
when 'f' is a pixel image.
o slrm
In the default case (where dataAtPoints is not given)
all spatial covariates, including the spatial coordinates x and y,
are now evaluated at the centre of each pixel. This improves
consistency with other implementations of spatial logistic regression.
o slrm
Silently ignores any arguments '...' that are not recognised by 'as.mask'
o summary.mppm
Improved summary of the dependence of the
interpoint interaction on the covariates.
o densityEqualSplit
New arguments 'at' and 'leaveoneout' for consistency with other functions.
o pairs.im
New argument 'drop'.
o distmap.psp
New arguments 'extras' and 'clip'
o discretise
New argument 'move.points' determines whether the point coordinates
are also discretised.
o summary.im
Output improved when the image is empty (i.e. when all pixel values
are undefined).
o rhohat
New option (smoother='piecewise') computes a piecewise-constant
estimate of rho(z).
o rhohat
The result now includes the 'average' intensity rho.
o distcdf
Arguments which are NULL will be treated as missing.
o distcdf
New argument 'savedenom'.
o densityHeat
The function formerly known as 'densityHeat' or 'densityHeatlpp'
is now renamed 'densityHeat.lpp' and is a method for the generic
'densityHeat'.
BUG FIXES
o nncross.ppp
If the argument 'by' was given, some of the results were incorrect.
[Spotted by Hank Stevens.]
Fixed.
o nncross.ppp, nncross.pp3
If 'iX' and 'iY' were given, some of the results were incorrect.
Fixed.
o density.lpp
The result had the wrong length if 'x' contained duplicated points
when 'weights' were given and 'at="points"'.
[Spotted by Andrea Gilardi]
Fixed.
o simulate.kppm
Conditional simulation crashed on rare occasions,
with an error about negative probabilities.
Fixed.
o model.matrix.mppm
If the model was fitted using 'gam', the resulting matrix
did not have an 'assign' attribute.
Fixed.
o model.depends
Crashed for models fitted using 'gam'.
Fixed.
o predict.slrm, fitted.slrm
Crashed if the model was fitted using split pixels (argument 'splitby').
Fixed.
o predict.slrm, fitted.slrm
Crashed in some cases when 'window' was given.
Fixed.
o update.slrm
Failed to find covariates that were provided in 'env'.
Fixed.
o cdf.test
Crashed if the covariate was constant.
Fixed.
CHANGES IN spatstat VERSION 2.1-0
OVERVIEW
o We thank Tilman Davies, Peter Diggle, Greg McSwiggan and Suman Rakshit
for contributions.
o diffusion kernel estimate of intensity
o New dataset 'btb'
o More support for spatial logistic regression models.
NEW FUNCTIONS
o densityHeat
New generic function for diffusion kernel estimation of intensity
o densityHeat.ppp
Diffusion kernel estimation of intensity in 2 dimensions
o densityHeat.lpp
Diffusion kernel estimation of intensity on a linear network
o slrm
'step' can now be applied to models fitted using 'slrm'.
NEW DATASETS
o btb
Bovine tuberculosis data, from Prof Peter Diggle.
CHANGES IN spatstat VERSION 2.0-1
OVERVIEW
o Minor changes to satisfy CRAN checks.
CHANGES IN spatstat VERSION 2.0-0
OVERVIEW
o We thank Corey Anderson, Michael Chirico, Andy Craig,
Marcelino de la Cruz, Tilman Davies, Pavel Fibich,
Kurt Hornik, Gopalan Nair, Yonatan Rosen and Rasmus Waagepetersen
for contributions.
o spatstat has been divided into 7 sub-packages
(spatstat.utils, spatstat.data, spatstat.sparse,
spatstat.geom, spatstat.core, spatstat.linnet and spatstat).
o Important bug fix in simulation of log-Gaussian Cox processes.
o Increased speed for large datasets.
o variance calculations handle larger datasets.
o predict.mppm now works for multitype point process models.
o Improved handling of 'newdata' in predict.mppm.
o More support for multi-dimensional patterns.
o Changed default value of 'stringsAsFactors'.
o spatstat now depends on R version 3.5.0 or later.
o spatstat now requires spatstat.utils version >= 1.18-0
o spatstat now requires spatstat.data version >= 1.7-0
o Bug fixes and minor improvements.
o Version nickname: "Caution: contains small parts"
NEW FUNCTIONS
o intersect.boxx
Compute intersection of boxes in multi-dimensional space
o scale.boxx, scale.ppx
Methods for 'scale' for boxes and patterns in multi-dimensional space
o shift.boxx, shift.ppx
Methods for 'shift' for boxes and patterns in multi-dimensional space
o is.boxx
Determine whether an object is a multidimensional box
SIGNIFICANT USER-VISIBLE CHANGES
o package structure
The original 'spatstat' package has been divided into 7 sub-packages
(spatstat.utils, spatstat.data, spatstat.sparse,
spatstat.geom, spatstat.core, spatstat.linnet and spatstat).
The remaining 'spatstat' package requires all the other sub-packages.
Your existing code scripts should still work with minimal changes.
o overall speed
Changes have been made to the internal code of spatstat
which should accelerate computations involving large datasets.
o vcov.ppm, summary.ppm
Variance calculations now handle larger datasets
because they use sparse arrays, by default.
o dirichletEdges
New argument 'clip'.
o rSSI
Accelerated.
o localpcf, localpcfinhom
New argument 'rvalue'.
o harmonise.im
The result belongs to class 'solist' and 'imlist'
so that it can be plotted.
o hyperframe, as.im.function
The formal default value of 'stringsAsFactors' has been changed
to 'NULL' to conform to changes in R. (The actual default value
is TRUE for R < 4.1.0 and FALSE for R >= 4.1.0)
o predict.mppm
Now supports multitype point process models.
o predict.mppm
Improved handling of argument 'newdata'
o densityHeat
Default behaviour has changed slightly.
New argument 'finespacing'.
o density.lpp
Accelerated when the pattern contains duplicated points.
o rotmean
The result now has the same 'unitname' as the input object X.
New argument 'adjust' controls the smoothing bandwidth.
o sessionInfo
Output now lists packages that are imported but not loaded.
o rlabel
New argument 'group' specifies that the points are divided into
several groups, and that relabelling is applied within each group.
o plot.psp
The code for 'style="width"' has been completely rewritten,
so that it no longer depends on plot.linim, and is more efficient.
The formal argument list has been extended.
o mincontrast
New argument 'action.bad.values' specifies what action is taken
when the summary function produces NA or NaN or infinite values.
o sessionLibs
Package names are now sorted alphabetically.
o [.linim
Accelerated.
o integral.im
Accelerated in the case where 'domain' is a tessellation.
o cbind.hyperframe
Row names are not altered (previously they were altered
using 'make.names')
o simulate.ppm
Now recognises the argument 'window' as an alternative to 'w'.
o kppm
Improved numerical robustness.
o Kcross, Gcross, Jcross
Function labels (shown on the plot legend) have been
improved when i = j.
o anova.mppm
Issues a warning when applied to random-effects models
(models fitted using the argument 'random').
o [.ppx
New argument 'clip'
BUG FIXES
o rLGCP, simulate.kppm
Simulation results for log-Gaussian Cox processes were incorrect
unless the pixel dimensions and pixel spacings were identical
on the horizontal and vertical axes. (If pixel dimensions were not
specified, then the results were incorrect whenever the Frame of the
simulation window was not a square.)
[Spotted by Tilman Davies.]
Fixed.
o crossdist.pp3
Results with periodic=TRUE were partially incorrect.
Fixed.
o deviance.lppm, pseudoR2.lppm
Results were completely incorrect, due to a coding error.
Fixed.
o colourmap
If a colour map was applied to numbers lying outside the range of the
colour map, the wrong number of NA's was sometimes produced.
Fixed.
o Gest
If correction="rs" or correction="km", then both the reduced-sample
(border correction) and Kaplan-Meier corrected estimates were calculated.
[Spotted by Gopalan Nair.]
Fixed.
o Lcross.inhom, Kcross.inhom, Kmulti.inhom
The option 'correction="none"' was accepted but ignored.
[Spotted by Corey Anderson.]
Fixed.
o rMatClust, rThomas, rCauchy, rVarGamma
If the fitted model was effectively a Poisson process,
the result did not have attributes 'Lambda' and 'parents'
even when the user requested them.
Fixed.
o affine.owin
For mask windows, the pixel resolution of the result
was too fine, leading to very large datasets.
Fixed.
o affine.im
If the transformation matrix was not diagonal, the pixel resolution
of the result was too fine, leading to very large datasets.
Fixed.
o plot.ppp
For a point pattern in a binary mask window,
if both arguments 'col' and 'cols' were given,
the points were coloured according to 'col', which was incorrect.
Fixed.
o dirichletEdges
Crashed if any edges crossed the boundary of the window.
Fixed.
o Vmark
Crashed if normalise=TRUE when there was only one column of marks.
(Spotted by Pavel Fibich.)
Fixed.
o unitname
Spatial datasets with incorrect internal format
(or using an out-of-date version of the spatstat format)
caused an error if the 'units' package was loaded.
Fixed.
o nnclean
Crashed if k >= npoints(X).
Fixed.
o print.ppm
Crashed sometimes when applied to the result of subfits().
Fixed.
o model.matrix.mppm
Crashed with random-effects models.
Fixed.
o anova.mppm
Crashed with random-effects models.
Fixed.
o objsurf.kppm
Crashed if the model was fitted by Palm likelihood (method="palm")
or second order composite likelihood (method="clik2").
Fixed.
o MinkowskiSum
Crashed sometimes with an error message about 'sumconnected'.
Fixed.
o simulate.rhohat
Crashed when applied to rhohat objects computed from data
on a linear network.
Fixed.
o hyperframe
Crashed if the argument 'row.names' was given
and the hyperframe had exactly one row.
Fixed.
CHANGES IN spatstat VERSION 1.64-1
OVERVIEW
o Important bug fix in vcov.ppm
o Relative risk estimation may include case weights.
o We thank Ian Buller, Brian Ripley, Maximilian Vogtland
and Maximilian Hesselbarth for contributions.
o Nickname: 'Help you I can, yes!'
SIGNIFICANT USER-VISIBLE CHANGES
o rshift.ppp, rshift.splitppp
New argument 'nsim'.
o relrisk.ppp
New argument 'weights'.
o density.splitppp
New argument 'weights'.
BUG FIXES
o vcov.ppm
Variances were sometimes overestimated for Gibbs models.
That is, entries of the Fisher information matrix were underestimated,
because some contributions due to interaction were omitted
(due to a coding error).
Fixed.
o density.ppp
Crashed when se=TRUE if there were multiple columns of 'weights'.
Fixed.
o rbind.hyperframe
Crashed unless all arguments had the same number of rows.
(Spotted by Maximilian Vogtland).
Fixed.
CHANGES IN spatstat VERSION 1.64-0
OVERVIEW
o We thank Robert Aue, Tilman Davies, Greg McSwiggan, Tyler Rudolph
and Rasmus Plenge Waagepetersen for contributions.
o Interactive graphics functions have been removed to a separate package.
o spatstat no longer needs the packages 'tcltk' and 'rpanel'.
o The suggested package 'maptools' should be version 0.9-9 or later.
o Important bug fix in density.ppp.
o Add new vertices to a linear network.
o Relative risk estimation on a network.
o Leave-one-out density estimation on a network.
o Improvements and extensions to linear networks code.
o Improvements to 'nndist' methods.
o Function lengths.psp has been renamed lengths_psp.
o Bug fixes related to mppm.
o Stability improvements.
o Version nickname: 'Susana Distancia'
NEW FUNCTIONS
o relrisk.lpp
Method for 'relrisk' for point patterns on a linear network.
o bw.relrisklpp
Bandwidth selection for relative risk on a network.
o densityfun.lpp
Method for 'densityfun' for point patterns on a linear network.
o addVertices
Add new vertices to a network, at locations outside the existing network.
o lengths_psp
This is the new name of the function 'lengths.psp',
which had to be changed because of a conflict with the generic 'lengths'.
o densityEqualSplit
The equal-split algorithm for kernel density estimation on a network
is now visible as a separate function.
o densityHeat
The heat-equation algorithm for kernel density estimation on a network
is now visible as a separate function. It has also been extended
to computing leave-one-out density estimates at the data points.
o hotrod
Compute the heat kernel kappa(u,v) on a one-dimensional line segment.
o heatkernelapprox
Calculate an approximation to the value of the heat kernel
on a network evaluated at the source point, kappa(u,u).
SIGNIFICANT USER-VISIBLE CHANGES
o nndist.pp3, nndist.ppx, nndist.lpp
These functions now recognise the argument 'by'
allowing computation of the nearest distance to each group of points.
o pairdist.lpp, crossdist.lpp
These functions can now handle large networks,
using the sparse representation.
o density.lpp, densityQuick.lpp
Infinite bandwidth (sigma=Inf) is now permitted,
and results in a density estimate that is constant over all locations.
o as.linnet.psp
The resulting network now has an attribute 'camefrom'
indicating the provenance of each line segment in the network.
o as.linnet.linnet
New argument 'maxsize'.
o repairNetwork
Increased capability of detecting and repairing inconsistencies.
o joinVertices
New argument 'marks'.
o insertVertices
Marks attached to the lines of the network are now retained.
o as.lpp
Accepts more data formats.
o iplot, iplot.ppp, iplot.layered, iplot.linnet, iplot.default
These interactive plotting functions have been removed from spatstat
into a new package 'spatstat.gui'
o istat
This interactive analysis function has been removed from spatstat
into a new package 'spatstat.gui'
o crossdist.lpp
New argument 'check'.
o lengths.psp
This function will soon be Deprecated,
in favour of the new name 'lengths_psp'
o density.lpp
Formal arguments changed. No effect on usage.
o integral.linim
Now handles complex-valued functions.
o transect.im
New argument 'nsample'.
o bw.lppl
Accelerated when distance="path".
o collapse.fv
Recognises the abbreviations used by fvnames()
BUG FIXES
o density.ppp
Edge correction factors were calculated incorrectly when the
window was not a rectangle, causing a negative bias in the
estimated intensity. [Spotted by Tilman Davies.]
Bug introduced in spatstat 1.57-0, october 2018.
Fixed.
o mppm
Internal data were malformed if the interaction was Hardcore()
or MultiHard() or a hybrid involving these interactions.
This caused various errors when the fitted model was used.
Fixed.
o mppm
Ignored the arguments 'nd' and 'eps' controlling the quadrature scheme.
Fixed.
o "[.linnet", "[.lpp"
In X[W] where W is a window, if a vertex of the network
lay exactly on the boundary of W, an edge of length zero was created.
Fixed.
o valid.ppm
Crashed sometimes when applied to the result of subfits().
Fixed.
o as.im.densityfun
Crashed if argument W was missing.
Fixed.
o as.linnet.linnet
This code could crash the R session, when sparse=FALSE, if there was
insufficient memory available to create the matrix of distances
between all pairs of network vertices.
Fixed.
o Summary.linim
A spurious warning was generated when the operation any() or all()
was applied to a logical-valued image on a network.
Fixed.
o "[<-.linim"
Crashed if the assignment would have replaced some existing NA values.
Fixed.
CHANGES IN spatstat VERSION 1.63-3
OVERVIEW
o Minor changes for compatibility with future versions of R
o Minor improvements
o Version nickname: "Wet paint"
SIGNIFICANT USER-VISIBLE CHANGES
o plot.ppp
The coordinate axes will be plotted if axes=TRUE.
Axis labels xlab, ylab will be plotted if ann=TRUE.
Comprehensive open-source toolbox for analysing Spatial Point
Patterns. Focused mainly on two-dimensional point patterns, including
multitype/marked points, in any spatial region. Also supports
three-dimensional point patterns, space-time point patterns in any
number of dimensions, point patterns on a linear network, and patterns
of other geometrical objects. Supports spatial covariate data such as
pixel images. Contains over 2000 functions for plotting spatial data,
exploratory data analysis, model-fitting, simulation, spatial
sampling, model diagnostics, and formal inference. Many data types and
exploratory methods are supported. Formal hypothesis tests of random
pattern and tests for covariate effects are also supported. Parametric
models can be fitted to point pattern data using the functions ppm(),
kppm(), slrm(), dppm() similar to glm(). Types of models include
Poisson, Gibbs and Cox point processes, Neyman-Scott cluster
processes, and determinantal point processes. Models may involve
dependence on covariates, inter-point interaction, cluster formation
and dependence on marks. Models are fitted by maximum likelihood,
logistic regression, minimum contrast, and composite likelihood
methods. A model can be fitted to a list of point patterns (replicated
point pattern data) using the function mppm(). The model can include
random effects and fixed effects depending on the experimental design,
in addition to all the features listed above. Fitted point process
models can be simulated, automatically. Formal hypothesis tests of a
fitted model are supported along with basic tools for model selection.
nia
mef
brook