Spatstat

analysing spatial point patterns

• Home
• News
• Download
• Docs
• Book
• Courses
• Help
• FAQ
• About

RELEASE NOTES

spatstat 1.42-0

27 May 2015

We thank Anders Bilgrau, Ute Hahn, Jack Hywood, Tuomas Rajala, Cody Schank, Olivia Semboli and Ben Taylor for contributions.

OVERVIEW

• Version nickname: ‘Life, The Universe and Everything’

• Permutation test for difference between groups of point patterns.

• Variational Bayes estimation for point process models.

• Score test in anova.ppm

• ROC curve, and discrimination index AUC, for fitted models.

• Interactive text editor for spatial datasets.

• Tools for analysing data on a tree.

• Kernel density/intensity estimation on a linear network.

• Random pixel noise.

• Improved behaviour of polygon geometry operations.

• Improved support for cluster and Cox models.

• Improved basic support for owin objects.

• Improved support for tessellations.

• More hierarchical Gibbs interactions.

• Modifications to Kest.

• summary method for Cox and cluster models.

• class ‘listof’ is almost completely replaced by ‘anylist’ and ‘solist’.

• Improvements and bug fixes.

• spatstat now depends on R version 3.2.0 or later.

NEW FUNCTIONS

• studpermu.test
  Studentised permutation test for difference between groups of point patterns.
  Generously contributed by Ute Hahn.

• AIC.kppm, extractAIC.kppm, logLik.kppm, nobs.kppm
  Methods for computing AIC for fitted Cox and cluster models.

• transmat
  Convert pixel arrays between different display conventions.

• roc
  Receiver Operating Characteristic curve.

• auc
  Discrimination index AUC (area under the ROC curve)

• edit.ppp, edit.psp, edit.im
  Interactive text editor works for spatial datasets.

• edit.hyperframe
  Interactive text editor works for hyperframes.

• parameters
  Extract all parameters from a fitted model.

• density.lpp
  Kernel estimation of point process intensity on a linear network.

• extractbranch, deletebranch, treeprune, treebranchlabels, begins
  Tools for analysing data on a tree.

• rnoise
  Random pixel noise.

• as.data.frame.owin
  Convert a window to a data frame.

• harmonise.owin
  Convert several binary mask windows to a common pixel grid.

• copyExampleFiles
  Copy the raw data files from an installed dataset
  to a chosen folder, for use in a practice exercise.

• density.ppplist
  Method for ‘density’ for lists of point patterns.

• inradius
  Radius of largest circle inside a window.

• mergeLevels
  Merge different levels of a factor.

• relevel.im, relevel.ppp, relevel.ppx
  Change the reference level of a factor.

• simulate.profilepl
  simulation method for models fitted by profile maximum pseudolikelihood.

• predict.rho2hat
  Prediction method for class rho2hat

• with.msr
  Evaluate (an expression involving) components of a measure.

• summary.kppm, print.summary.kppm, coef.summary.kppm
  Methods for ‘summary’ and ‘coef(summary(..))’ for Cox and cluster models.

• as.im.funxy
  Method for as.im for class funxy.

• shift.linim, scalardilate.linim, affine.linim
  Geometrical transformations for ‘linim’ objects.

• Smooth.solist
  Smooth method for a list of spatial objects.

• unitname.tess, unitname<-.tess
  Tessellations now keep track of the name of the unit of length.

• dirichletAreas
  Faster algorithm for tile.areas(dirichlet(X)).

• identify.lpp
  Method for ‘identify’ for point patterns on a linear network.

• HierStraussHard, HierHard
  Hierarchical interactions for Gibbs models.

• delaunayDistance, delaunayNetwork, dirichletEdges,
  dirichletNetwork, dirichletVertices, dirichletWeights
  These functions will replace delaunay.distance, delaunay.network,
  dirichlet.edges, dirichlet.network, dirichlet.vertices and
  dirichlet.weights respectively.
  The latter are now ‘deprecated’.

SIGNIFICANT USER-VISIBLE CHANGES

• ppm
  Now supports Variational Bayes fitting method.

• kppm
  ‘AIC’ and ‘step’ now work for kppm objects
  fitted using maximum Palm likelihood.

• kppm
  The default for the weight function ‘weightfun’ has been changed,
  for better performance.

• envelope
  envelope methods now have argument ‘use.theory’ specifying whether
  to use the ‘theoretical’ value of the summary function when constructing
  simultaneous envelopes.

• anova.ppm
  Now performs the Score Test, for Poisson models only,
  if argument test=”Rao” or test=”score”.

• Kest
  New argument ‘rmax’ controls maximum value of argument ‘r’

• diagnose.ppm
  Now computes and prints the null standard deviation of the
  smoothed Pearson residual field, when appropriate.

• nncorr, nnmean, nnvario
  New argument ‘k’ specifies k-th nearest neighbour.

• quadrat.test.ppp, quadrat.test.quadratcount
  New argument ‘lambda’ supports a test of the Poisson process
  with given intensity ‘lambda’.

• clickpoly, clickbox
  These functions now handle graphical arguments to polygon()
  when drawing the resulting polygon or rectangle.

• owin, as.owin, as.mask
  owin(mask=D) or as.owin(D) or as.mask(D) will produce a binary mask window
  if D is a data frame with two columns of (x,y) coordinates
  or a data frame with three columns containing (x,y,logical).

• as.owin.data.frame
  W can now be a data frame with only two columns,
  giving the spatial coordinates of the pixels that are inside the window.

• rose
  Tick marks now have labels showing the angle (in degrees or radians).

• distcdf
  New argument ‘regularise’ determines whether values at short distances
  will be smoothed to avoid discretisation artefacts.

• rpoislinetess
  Return value now has an attribute ‘lines’ giving the
  realisation of the Poisson line process.

• intersect.owin, union.owin, setminus.owin
  New argument ‘p’ controls resolution of polygon clipping algorithm.

• intersect.owin, union.owin
  Arguments may be lists of windows, of class ‘solist’.
  Formal arguments A and B have been removed.

• superimpose
  Now handles lists of point patterns
  (objects of class ‘ppplist’ or ‘splitppp’)

• density.ppp
  New argument ‘positive’ allows the user to stipulate that
  density values must be positive (avoiding numerical errors
  which occasionally produce small negative values).

• adaptive.density
  Now accepts f = 0 (uniform intensity estimate) and
  f = 1 (Voronoi intensity estimate) as well as 0 < f < 1.
  Algorithm accelerated.

• rSSI
  Can now generate inhomogeneous patterns.

• effectfun
  Now works for ‘kppm’ and ‘lppm’ objects as well.

• integral.im, integral.msr
  Argument ‘domain’ can now be a tessellation;
  the integral over each tile of the tessellation is returned.

• allstats, compareFit, markcorr, split.ppx, by.ppp
  Result is now of class ‘anylist’.

• by.im, density.splitppp, idw, model.images, nnmark, pixellate.ppp,
  predict.lppm, predict.ppm, quadratcount.splitppp, quadratresample, relrisk,
  Smooth.msr, split.im, tiles
  Result is now of class ‘solist’.

• split.ppp
  New argument ‘reduce’.
  Result now inherits class ‘ppplist’ and ‘solist’, as well as ‘splitppp’

• rLGCP
  New argument ‘nsim’ allows multiple patterns to be generated.

• alltypes
  New argument ‘reuse’ determines whether all simulation envelopes
  are based on the same set of simulated patterns, or on independent sets.

• rpoispp, runifpoint
  New argument ‘ex’ makes it possible to generate a random pattern
  similar to an example point pattern.

• effectfun
  Argument ‘covname’ is not needed if the model has only one covariate.

• quadratcount
  Argument ‘tess’ can now be anything acceptable to as.tess.

• tess
  New argument ‘unitname’ specifies the name of the unit of length.
  If it is missing, unitname information will be extracted from the
  other data.

• intersect.tess, chop.tess, quadrats
  Results of these functions now have the same ‘unitname’ as their input.

• persp.im, nnclean, plot.qqppm, plot.bw.optim
  These plotting functions now obey spatstat.options(‘monochrome’)

• lurking
  Now returns an object of class ‘lurk’ which has a plot method.
  Two-standard-deviation limits are now plotted using grey shading.

• marktable
  New argument ‘N’ for studying the N nearest neighbours.
  New argument ‘collapse’ for manipulating the contingency table.

• harmonise.fv
  Now discards columns with names which do not match.

• eval.fv
  New argument ‘equiv’ can be used to declare that two columns
  with different names in different objects are equivalent.

• quantile.ewcdf
  New argument ‘type’ controls the type of quantile.

• plot.imlist
  New argument ‘plotcommand’ specifies how to plot each image.

• persp.im
  The lower extent of the apron can now be controlled by ‘zlim’.

• quadscheme
  Argument ‘method’ is partially matched.

• Kdot, Ldot
  New argument ‘from’ is an alternative to ‘i’.

• Kcross, Lcross
  New arguments ‘from’ and ‘to’ are alternatives to ‘i’ and ‘j’ respectively.

• varblock
  Changed the ordering (and therefore default colours/styles)
  of curves in the plot, to match other functions like lohboot.

• bw.diggle
  New argument ‘nr’ controls accuracy.

• textureplot
  Now accepts a pixel image, a tessellation, or anything acceptable to as.tess.

• textureplot
  Line spacing in legend now matches line spacing in main display.

• [.tess
  Subset index can now be a window.

• plot.tess
  Can now plot a text label in each tile.

• plot.tess
  New argument ‘do.plot’.

• MultiHard, MultiStrauss, MultiStraussHard, HierStrauss
  Printed output of fitted model now respects spatstat.options(‘terse’).

• print.ppm
  Reduced redundancy in output in some cases.

• print.msr
  Responds better to spatstat.options(‘terse’).

• print.ppm, print.fii, print.interact
  Irregular parameters are now printed to the number of significant figures
  specified by options(“digits”).

• square
  New argument ‘unitname’.

• plot.fv
  Return value is now invisible.

• delaunay.distance, delaunay.network, dirichlet.edges,
  dirichlet.network, dirichlet.vertices
  These functions are now ‘deprecated’, and will be replaced by
  delaunayDistance, delaunayNetwork, dirichletEdges,
  dirichletNetwork and dirichletVertices respectively.

• data(residualspaper)
  In the real datasets (Fig1 and Fig11), the name of the unit of length
  has now been recorded.

• rLGCP
  This function now requires the package ‘RandomFields’
  to be loaded explicitly by library(RandomFields) or require(RandomFields),
  unless model=”exp”.

• iplot, istat
  These functions now require the package ‘rpanel’
  to be loaded explicitly by library(rpanel) or require(rpanel).

• ppm, quadscheme
  Improved calculation of Dirichlet weights.

• countends
  New argument ‘toler’ controls numerical errors

• diagnose.ppm
  Improved handling of additional graphics arguments.

• pcf3est
  Mathematical labels changed.

• plot.hyperframe
  Default margin spacing has been increased.

BUG FIXES

1. Kinhom, Linhom
   The value of ‘theo’ was erroneously rescaled
   by a small amount, when renormalise=TRUE (the default).
   Fixed.

2. Kmark
   Values were erroneously rescaled.
   Fixed.

3. union.owin
   Strange results were sometimes obtained when taking the union
   of more than two windows.
   Fixed.

4. rpoispp3
   Implementation was incorrect for nsim > 1.
   (Results may have been incorrect.)
   Spotted by Jack Hywood.
   Fixed.

5. as.owin.data.frame
   Crashed if the window was not connected.
   Fixed.

6. Frame<-
   Crashed when applied to a binary mask.
   Fixed.

7. rho2hat
   Crashed if cov1=”x” and cov2=”y”.
   Fixed.

8. as.mask
   Crashed sometimes when only the argument ‘xy’ was given.
   Fixed.

9. ppm
   Crashed (rarely) when method=’ho’ if the simulated pattern was empty.
   Fixed.

10. istat, iplot
    Crashed in recent versions of rpanel.
    Fixed.

11. convexhull
    Crashed if applied to a ‘psp’ object.
    Fixed.

12. plot.ppm
    Crashed with message about ‘variable lengths differ’.
    Fixed.

13. plot.solist
    Crashed when applied to a list of point patterns
    if some patterns had more than one column of marks.
    Fixed.

14. Smooth.ppp
    Crashed if applied to a point pattern with several columns of marks
    if some of the columns were factors.
    Fixed.

15. runifpoint3, rpoispp3
    Crashed if nsim > 1.
    Spotted by Jack Hywood.
    Fixed.

16. hist.im
    Crashed if argument ‘freq’ was given.
    Fixed.

17. MultiStraussHard
    Generated misleading error messages (e.g. ‘model is invalid’)
    when arguments ‘iradii’ and ‘hradii’ did not have the same pattern of NA’s.
    Fixed.

18. plot.solist
    Figures were sometimes aligned incorrectly
    when the argument ‘panel.args’ was given.
    Fixed.

19. scaletointerval
    Results sometimes fell slightly outside the desired interval
    due to numerical error.
    Fixed.

20. plot.solist
    Behaved incorrectly when plotcommand=’persp’.
    Fixed.

21. ”[.hyperframe”
    Sometimes returned an ‘anylist’ when it should have returned a ‘solist’.
    Fixed.

22. plot.im
    Did not plot surrounding frame box when ribbon=FALSE.
    Fixed.

23. envelope
    The functions stored when savefuns=TRUE did not inherit
    the correct name for the unit of length.
    Fixed.

24. print.ppm, print.fii, print.interact
    Layout was misaligned.
    Fixed.

25. plot.plotppm
    Paused for input when it was not appropriate.
    Fixed.

26. plot.fv
    On png devices, the legend box was drawn with a white background,
    obscuring the main plot.
    Fixed.

27. plot.owin, plot.ppp, plot.im
    There was unnecessary extra space above the main title.
    Fixed.

28. plot.rho2hat
    Colour map ribbon was drawn but not annotated.
    Fixed.

29. density.splitppp, density.ppplist
    Format was out of order if se=TRUE.
    Fixed.

30. MultiStraussHard
    project.ppm sometimes yielded a model that was still invalid.
    Fixed.

Release notes are available in raw text format here.