NEWS

        CHANGES IN spatstat.model VERSION 3.7-2

OVERVIEW

    o Package check time reduced, to satisfy CRAN requirements.

    o Variance-covariance matrix for determinantal models.
    
    o More diagnostics for determinantal models.

    o Added variable plots for all kinds of models.
    
    o Minor improvements and bug fixes.

NEW FUNCTIONS

    o auc.dppm
    Area under the ROC curve for a fitted determinantal point process model.
    Method for 'auc' for objects of class 'dppm'.
    
    o parres.dppm
    Partial residuals for a fitted determinantal point process model.
    Method for 'parres' for objects of class 'dppm'.

    o vcov.dppm
    Variance-covariance matrix for the trend parameters in a
    fitted determinantal point process model.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o addvar
    Now accepts point process models of class ppm, kppm, dppm, rppm, lppm.

    o addvar
    For non-Poisson models, significance bands are computed by simulation.

    o plot.kppm
    For models fitted by Palm likelihood or composite likelihood, the plot
    selected by 'what="statistic"' displays the pair correlation function.

    o rhohat.ppm, rhohat.slrm
    Methods for 'rhohat' now recognise the argument 'delta'
    as an alternative to argument 'breaks' when smoother="piecewise".

    o palmdiagnose
    Now accepts objects of class 'dppm' as well as 'kppm'.
    
    o palmdiagnose
    New argument 'delta' as an alternative to 'breaks'.

    o objsurf.dppm, objsurf.kppm, objsurf.minconfit
    Arguments 'xlim, ylim' renamed 'par1lim, par2lim'.
    
BUG FIXES

   o plot.kppm
   For inhomogeneous models fitted by minimum contrast,
   a warning was issued about truncated function labels.
   Fixed.
   
   o plot.minconfit
   In some cases, a warning was issued about truncated function labels.
   Fixed.

   o palmdiagnose
   Crashed sometimes with a message about a missing polygon.
   Fixed.

        CHANGES IN spatstat.model VERSION 3.7-1

OVERVIEW

    o We thank Martin Emons for contributions.

    o More methods for 'intensity' and 'repul'.

    o Improved handling of boundary pixels in model predictions.
    
    o Bug fixes and minor improvements.

NEW FUNCTIONS

    o repul.detpointprocfamily
    Repulsion index for determinantal point process models.
    A new method for the generic 'repul'.

    o intensity.kppm
    Method for 'intensity' for fitted cluster processes and Cox processes.

    o intensity.mppm
    Method for 'intensity' for models fitted to multiple point patterns.

    o repul.kppm, repul.ppm, repul.slrm, repul.mppm, repul.rppm, repul.clusterprocess
    New methods for the generic 'repul' calculate the repulsion index
    for Poisson, Gibbs, Cox and Neyman-Scott point process models.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o predict.ppm, predict.kppm, predict.dppm, predict.rppm, predict.mppm
    These methods now recognise argument 'rule.pix' passed to owin2mask.
    When the original point pattern dataset was observed in a polygonal window,
    setting rule.pix="cover" ensures that pixels which partially intersect
    the polygon will be included in the resulting pixel image.

    o relrisk.ppm
    Now recognises argument 'rule.pix' passed to owin2mask.
    When the original point pattern dataset was observed in a polygonal window,
    setting rule.pix="cover" ensures that pixels which partially intersect
    the polygon will be included in the resulting pixel image.

    o Kcom, Kres
    Computation accelerated when using Ripley isotropic correction.

    o print.kppm
    The index of clustering strength is now printed when 'x' is
    a log-Gaussian Cox process (as well as when 'x' is a Neyman-Scott
    cluster process).
    
BUG FIXES

   o vcov.ppm
   There was a loophole in the code which *could* have caused
   the result of vcov.ppm to be incorrect for pairwise interaction models
   in a very rare case. No examples have been found.
   Fixed.

   o mppm
   Generated an error about a subscript out of bounds,
   if the package 'formula.tools' was loaded.
   [Spotted by Martin Emons.]
   Fixed.

   o pcfmodel.ppm
   Crashed for pairwise interaction models with a hard core component.
   Fixed.
   
        CHANGES IN spatstat.model VERSION 3.7-0

OVERVIEW

    o New generic 'Lmodel'.

    o Methods for class 'clustermodel'

    o 'envelope' method for 'dppm'

    o More options for 'simulate' argument in 'envelope'

    o Specify the 'theoretical' value of the summary function in 'envelope'

NEW FUNCTIONS

    o envelope.dppm
    Method for 'envelope' for fitted models of class 'dppm'
   
    o Lmodel, Lmodel.kppm, Lmodel.dppm, Lmodel.ppm, Lmodel.detpointprocfamily
    Given a point process model, return a function in the R language
    which computes the theoretical L-function of the model.

    o Kmodel.clustermodel, pcfmodel.clustermodel, Lmodel.clustermodel
    Calculate the theoretical K-function, L-function or pair correlation
    function for a model of class 'clustermodel'.
   
SIGNIFICANT USER-VISIBLE CHANGES

    o psib, persist, clusterstrength, varcount
    These functions now accept an object of class 'clustermodel'.

    o envelope.ppm, envelope.kppm, envelope.slrm
    Argument 'simulate' can be a point process model of class
    'ppm', 'kppm', 'dppm', 'slrm', 'clusterprocess' or 'detpointprocfamily'.

    o envelope.ppm, envelope.kppm, envelope.slrm
    New argument 'theoryfun' can be used to provide the theoretical value
    of the summary function.
    
        CHANGES IN spatstat.model VERSION 3.6-1

OVERVIEW

    o Minor changes to satisfy CRAN.
    

        CHANGES IN spatstat.model VERSION 3.6-0

OVERVIEW

    o We thank Gabriela Calana Somoza, Bethany Macdonald
    and Rasmus Waagepetersen for contributions.
    
    o Waagepetersen's composite likelihood for cluster processes.
    
    o Convert a recursively partitioned point process model to a tessellation.

    o Improvements to 'profilepl' and 'effectfun'.

    o Diagnostic functions have argument 'rmax'.
    
    o Cluster strength index and spatial persistence index.

    o Bug fixes and minor improvements.
    
NEW FUNCTIONS

    o as.tess.rppm
    Convert a recursively partitioned point process model
    to a spatial tessellation.

    o clusterstrength
    Cluster strength index for a cluster process or Cox process.

    o formula.rppm
    Extract the model formula for a recursively partitioned point process model.

    o persist
    Spatial persistence index for a cluster process.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o kppm
    New option: method="waag" fits the model by maximising
    Waagepetersen's (2007) second order composite likelihood
    [Suggested by Bethany Macdonald]

    o effectfun
    Improved defaults for a recursive partition model (class 'rppm').
    The user only needs to provide values for those covariates which
    are used in the decision tree. If the entire tree involves only one
    variable, then 'covname' is not needed.

    o effectfun
    If a spatial covariate is constant, its value does not need to be
    provided in the call to effectfun().

    o as.fv.kppm
    Improved arguments for empirical estimation of pair correlation function
    when the model was fitted with a non-default value of 'method'.

    o Gcom, Kcom, psstA, psstG, psst
    These diagnostic functions have a new formal argument 'rmax'.

    o Gres, Kres
    These diagnostic functions accept argument 'rmax'.
    
    o lurking.ppm, lurking.ppp, lurking.slrm
    Argument 'covariate' can be a function(x,y)
    or one of the strings "x" or "y".

    o lurking.ppm, lurking.ppp, lurking.slrm, plot.lurk
    When the covariate is a cartesian coordinate or a distance function,
    the unit of length is stated on the horizontal axis label of the plot.

    o Penttinen
    Computation of interaction terms has been accelerated.
    
    o profilepl
    It is now possible to fit a model with a formula like ~ pmax(x, thresh)
    where 'thresh' is an irregular parameter.

    o Strauss
    Computation of interaction terms has been slightly accelerated.
    
BUG FIXES

   o effectfun
   Crashed if the model involved a covariate function
   with extra arguments passed through 'covfunargs'.
   Fixed.
   
   o lurking
   Crashed in rare cases when 'envelope=TRUE' with message 'Xsim not found'.
   [Spotted by Gabriela Calana Somoza.]
   Fixed.

   o plot.leverage.ppm
   Return value was NULL.
   Fixed.
   
   o update.ppm
   update(object, X) crashed sometimes if X was a point pattern
   whose observation window was not identical to the window of the
   original point pattern dataset.
   Fixed.

        CHANGES IN spatstat.model VERSION 3.5-0

OVERVIEW

    o Models for replicated point patterns on linear networks.
    
    o ROC curves for determinantal point process models.

    o Partial residuals for cluster process and Cox process models.
    
    o Extended facilities for spatial logistic regression.

    o Bug fixes and internal improvements.

NEW FUNCTIONS

    o diagnose
    New generic function for model diagnostics.

    o parres
    New generic function for partial residuals.
    
    o parres.ppm
    The original function 'parres' has now become the method 'parres.ppm'.

    o parres.kppm
    Partial residual plot for fitted cluster process or Cox process models.

    o roc.dppm
    Compute ROC curve for a determinantal point process model.

SIGNIFICANT USER-VISIBLE CHANGES

    o diagnose.ppm
    This is now a method for the generic function 'diagnose'.
    
    o fitted.slrm
    Argument 'leaveoneout' is no longer ignored when 'dataonly=FALSE';
    it now supports 'leave-one-pixel-out' prediction.

    o mppm
    Now allows the response variable to be a point pattern on a linear network.
    This is experimental code.

    o predict.slrm
    New argument 'leaveoneout' supports leave-one-pixel-out calculation
    of predicted values, and new argument 'fast' determines the algorithm used.

    o print.slrm
    Now prints information about the pixel raster.

    o residualMeasure
    More options for argument 'lambda'.
    
    o roc.slrm
    Changed interpretation of argument 'leaveoneout=TRUE'.
    This now performs 'leave-one-pixel-out' calculation
    (previously it performed 'leave-one-point-out' calculation).
    
    o slrm
    New argument 'ruleAtPoints' controls discretisation rule at data points.

    o slrm
    Arguments '...' are now passed to 'owin2mask' to determine the
    polygon-to-raster discretisation rule for a covariate of class 'owin'.

    o summary.slrm
    Now prints information about the pixel raster.
    
BUG FIXES

    o fitted.slrm
    Calculation was incorrect when leaveoneout=TRUE and type != "intensity".
    Fixed.
    
        CHANGES IN spatstat.model VERSION 3.4-2

OVERVIEW

    o Internal changes to satisfy CRAN.

        CHANGES IN spatstat.model VERSION 3.4-1

OVERVIEW

    o Improvements to intensity.ppm for Geyer models.

    o Simulation of a fitted model in a different window is better supported.

    o Minor improvements.
    
    o We thank Andrea Gilardi for contributions.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o intensity.ppm
    For Geyer models, the computation is faster,
    and supports more values of the saturation parameter 'sat',
    namely sat = 0.5, 1, 1.5, 2, 2.5 and 3.

    o ppm, kppm, dppm, slrm, mppm
    These model-fitting functions now handle the case where the
    point pattern is an 'NAobject'.

    o simulate.ppm, rmh.ppm, rmhmodel.ppm
    New argument 'newdata' allows the model to be simulated
    in a completely different window, or using different covariate values.
    
    o simulate.ppm, rmh.ppm, rmhmodel.ppm
    A warning or error occurs if there are too many NA values
    of the model trend in the desired window. 

BUG FIXES

    o predict.mppm
    Crashed if a covariate contained NA values and
    the model included random effects. 
    [Spotted by Andrea Gilardi.]
    Fixed.

        CHANGES IN spatstat.model VERSION 3.4-0

OVERVIEW

    o Extensions to ROC curves.

NEW FUNCTIONS

    o addROC
    Calculate partial ROC curve for adding a new covariate to a model.

    o dropROC
    Calculate partial ROC curve for removing a covariate from a model.

    o dropply, addapply
    Consider all single-variable additions or deletions in a model.

SIGNIFICANT USER-VISIBLE CHANGES

    o mppm
    Recognises argument 'quad.args'.
    
    o roc.ppm, roc.slrm, roc.kppm
    New argument 'baseline' allows calculation of ROC relative to a baseline.

    o roc.ppm, roc.slrm, roc.kppm
    New argument 'method' determines the estimation method.
    New options include a kernel smoothing estimate and a monotone estimate.
    
    o roc.ppm, roc.slrm, roc.kppm
    New arguments 'CI', 'alpha' for calculating confidence intervals.

    o roc.ppm, roc.slrm, roc.kppm
    New argument 'subset' specifies a sub-region of the spatial domain
    in which the ROC should be calculated.

    o roc.ppm, roc.slrm, roc.kppm
    New argument 'leaveoneout' specifies whether to use leave-one-out estimates
    of intensity at the data points.

        CHANGES IN spatstat.model VERSION 3.3-6

OVERVIEW

    o Simulation of 'zclustermodel' objects.
    
    o Documentation improvements and bug fixes.
    
NEW FUNCTIONS

    o simulate.zclustermodel
    Method for simulating objects of class 'zclustermodel'.
    
BUG FIXES

    o intensity.zclustermodel
    Always returned the value zero.
    Fixed.

        CHANGES IN spatstat.model VERSION 3.3-5

OVERVIEW

    o We thank John Ginos for contributions.
    
    o Bug fix.

BUG FIXES

    o simulate.kppm
    Argument saveLambda=TRUE was ignored for nonstationary LGCP models.
    [Spotted by John Ginos.]
    Fixed.

        CHANGES IN spatstat.model VERSION 3.3-4

OVERVIEW

    o Improvements to documentation.

    o Bug fix affecting spatstat.linnet

BUG FIXES

    o internal code
    For point process models on a linear network (class 'lppm')
    involving covariates of class 'lintess', the internal structure
    of the fitted model was corrupted, leading to errors in calculating
    properties of the fitted model, such as predict.lppm.
    [Spotted by Andrea Gilardi.]
    [Bug fix requires changes in spatstat.model internal code]
    Fixed.
    
        CHANGES IN spatstat.model VERSION 3.3-3

OVERVIEW

    o Bug fixes and minor improvements.

SIGNIFICANT USER-VISIBLE CHANGES

    o model.images.ppm
    Now recognises arguments passed to 'as.mask'
    to control the pixel raster for the images.
    
BUG FIXES

    o improve.kppm
    Crashed if NA's were present in the covariate values.
    Fixed.

    o model.matrix.ppm
    Crashed with a message about 'logical index too long',
    if NA's were present in the covariate values.
    Fixed.

        CHANGES IN spatstat.model VERSION 3.3-2

OVERVIEW

    o Tweaks to documentation.

    o Improved Palm diagnostic plot.

SIGNIFICANT USER-VISIBLE CHANGES

    o plot.palmdiag
    Improved placement of legend.
    
        CHANGES IN spatstat.model VERSION 3.3-1

OVERVIEW

    o Internal changes to satisfy CRAN package checker.
    

        CHANGES IN spatstat.model VERSION 3.3-0

OVERVIEW

    o Package now depends on 'spatstat.univar'.
    
    o Easier control over quadrature schemes.

    o More options in fitted.slrm
    
    o Bug fix in predict.ppm.

    o Internal improvements.

PACKAGE DEPENDENCE

    o spatstat.model now depends on the new package 'spatstat.univar'.

SIGNIFICANT USER-VISIBLE CHANGES

    o ppm.ppp
    New argument 'quad.args' is a list of arguments passed to 'quadscheme'
    to control the construction of the quadrature scheme.

    o fitted.slrm
    New argument 'type' allows calculation of fitted probabilities, intensities
    or link function values.

    o fitted.slrm
    New arguments 'dataonly' and 'leaveoneout' allow calculation of fitted
    values at the data points only, using leave-one-out calculation if desired.

BUG FIXES

    o predict.ppm
    Argument 'eps' was ignored in many cases.
    Fixed.

        CHANGES IN spatstat.model VERSION 3.2-11

OVERVIEW

    o Slightly accelerated.

SIGNIFICANT USER-VISIBLE CHANGES

    o spatstat.model package
    Some computations slightly accelerated.

        CHANGES IN spatstat.model VERSION 3.2-10

OVERVIEW

    o Internal bug fix.

        CHANGES IN spatstat.model VERSION 3.2-9

OVERVIEW

    o We thank Marta Luraschi for contributions.
    
    o Minor improvements and bug fixes.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o vcov.kppm
    If any quadrature points have NA values for one of the covariates,
    these quadrature points are ignored in the variance calculation,
    with a warning.
    
    o vcov.kppm
    Minor change to formal arguments.

    o vcov.ppm
    Minor change to formal arguments.

BUG FIXES

    o vcov.kppm
    If any quadrature points had NA values for one of the covariates,
    the result was a matrix of NA values.
    Fixed.
    
    
        CHANGES IN spatstat.model VERSION 3.2-8

OVERVIEW

    o spatstat.model no longer suggests package 'maptools' or 'RandomFields'.

    o code for fitting log-Gaussian Cox models has changed.
    
    o minor improvements to help files.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o kppm
    The code for fitting log-Gaussian Cox process models (clusters="LGCP")
    has been re-implemented without using the package 'RandomFields'.
    The current code supports the 'exponential', 'gauss', 'stable',
    'gencauchy' and 'matern' covariance models.
    
    o lgcp.estK, lgcp.estpcf
    This code for fitting log-Gaussian Cox process models
    has been re-implemented without using the package 'RandomFields'.
    The current code supports the 'exponential', 'gauss', 'stable',
    'gencauchy' and 'matern' covariance models.

    o simulate.kppm
    For log-Gaussian Cox process models (clusters='LGCP') the simulation
    algorithm has been completely re-implemented to avoid dependence
    on the defunct package 'RandomFields'. For details, see the help
    for 'rLGCP' in the package 'spatstat.random'.

        CHANGES IN spatstat.model VERSION 3.2-6

OVERVIEW

    o Internal improvements.

SIGNIFICANT USER-VISIBLE CHANGES

    o ppm
    Internal changes to support improvements in spatstat.linnet::lppm.

        CHANGES IN spatstat.model VERSION 3.2-4

OVERVIEW

    o Minor improvements.

SIGNIFICANT USER-VISIBLE CHANGES

    o cdf.test.ppm
    Recognises argument 'rule.eps' passed to 'as.mask'.

    o dfbetas.ppm
    Recognises argument 'rule.eps' passed to 'as.mask'.

    o leverage.ppm
    Recognises argument 'rule.eps' passed to 'as.mask'.

    o rhohat.ppm
    New argument 'rule.eps' passed to 'as.mask'.


        CHANGES IN spatstat.model VERSION 3.2-3

OVERVIEW

    o Tweak to satisfy package checker.


        CHANGES IN spatstat.model VERSION 3.2-2

OVERVIEW

    o Accelerated some code.

    o Standard errors are now available for 'ppm' models fitted using 'gam'.
    
    o Internal reorganisation.

    o Minor changes to documentation.

    o Bug fixes.

SIGNIFICANT USER-VISIBLE CHANGES

    o plot.mppm
    New argument 'main'.
    
BUG FIXES

    o predict.ppm
    Standard error calculation ('se=TRUE') crashed if the fitted model
    was a generalised additive model (fitted with 'use.gam=TRUE').
    Fixed.

    o effectfun
    Standard error calculation ('se.fit=TRUE') crashed if the fitted model
    was a generalised additive model (fitted with 'use.gam=TRUE').
    Fixed.

    o parres
    If 'model' was a large object, computation was extremely slow
    or terminated with a message about 'deparse'.
    Fixed.

    o plot.mppm
    If the fitted model 'x' was a large object, computation was extremely slow
    or terminated with a message about 'deparse'.
    Fixed.

    o predict.ppm
    If 'new.coef' was given and the fitted model 'object' was a large object,
    computation was extremely slow, or terminated with message about 'deparse'. 
    Fixed.
    
        CHANGES IN spatstat.model VERSION 3.2-1

OVERVIEW

    o Minor changes to citation file, to satisfy CRAN.
    

        CHANGES IN spatstat.model VERSION 3.2-0

OVERVIEW

    o Improvements to 'update' methods for point process models.

    o New 'update' methods for classes 'dppm' and 'rppm'.

    o Minor improvements and bug fixes.

NEW FUNCTIONS

    o update.dppm
    Update method for determinantal point process models.
    
    o update.rppm
    Update method for recursively partitioned point process models.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o update.slrm
    Now accepts the idiom 'update(object, X)' where X is a point pattern.

    o update.ppm, update.kppm, update.dppm, update.slrm, update.rppm
    All of these methods now accept the idiom 'update(object, X)'
    where X is a point pattern.

    o print.ppm, summary.ppm
    Prints the name of the point pattern dataset to which the model was fitted.

    o update.ppm
    Internal improvements.

        CHANGES IN spatstat.model VERSION 3.1-2

OVERVIEW

    o Minor changes to satisfy CRAN package checker.
    

        CHANGES IN spatstat.model VERSION 3.1-1

OVERVIEW

    o We thank Bethany Macdonald for contributions.
    
    o kppm has been accelerated when method='palm' or 'clik2'

    o kppm can save the history of the optimisation algorithm.

    o Bug fixes and internal tweaks.

SIGNIFICANT USER-VISIBLE CHANGES

    o kppm
    New argument 'trajectory' specifies whether to save the history
    of function evaluations performed by the optimization algorithm.

    o kppm
    Computation accelerated when 'method="palm"' or 'method='clik2'".
    [Kindly contributed by Bethany Macdonald.]
    
BUG FIXES

    o simulate.kppm
    If the model was very close to a Poisson process, and if saveLambda=TRUE was selected,
    the attribute "Lambda" was incorrectly labelled "lambda".
    Fixed.

    o simulate.kppm
    Simulation of the variance-gamma model terminated with an error about the value of 'nu' 
    (with recent versions of spatstat.random).
    Fixed.
    
    o kppm
    Terminated with an error about missing argument 'A', if penalised=TRUE.
    Fixed.

    o summary.kppm
    Did not correctly recognise when a model was fitted using a penalty.
    Fixed.
    
        CHANGES IN spatstat.model VERSION 3.1-0

OVERVIEW

    o We thank Tilman Davies and Martin Hazelton for contributions.

    o Penalised model-fitting for Neyman-Scott cluster process models.

    o Index of the strength of clustering in a Neyman-Scott cluster process model.
    
    o Probability of having any siblings.

    o More information is printed about Neyman-Scott cluster process models.
    
    o Minor improvements.

NEW CLASSES

    o traj
    Trajectory (history of function evaluations) in a model
    that was fitted by optimisation.
    
NEW FUNCTIONS

    o panysib
    Probability that a point in a cluster process has any siblings.

    o is.poissonclusterprocess
    Detects whether a given model is a Poisson cluster process
    (which includes Neyman-Scott processes).

    o traj, print.traj, plot.traj, lines.traj
    Extract, print and plot the trajectory of function evaluations.


SIGNIFICANT USER-VISIBLE CHANGES

    o kppm
    New argument 'penalised' supports penalised model-fitting with a penalty against
    extremely large or small values of the cluster scale.

    o print.kppm, summary.kppm
    Additional characteristics of the fitted model are reported,
    including the cluster strength 'phi' and the sibling probability.

    o varcount
    New argument 'relative' (supports calculation of the overdispersion index).

    o plot.palmdiag
    Improved calculation of y axis limits.
    
        CHANGES IN spatstat.model VERSION 3.0-3

OVERVIEW

    o Palm intensity diagnostic.

NEW FUNCTIONS

    o palmdiagnose, plot.palmdiag
    Palm intensity diagnostic plot for cluster process models
    proposed by Tanaka, Ogata and Stoyan.

        CHANGES IN spatstat.model VERSION 3.0-2

OVERVIEW

    o Minor changes to placate the package checker.


        CHANGES IN spatstat.model VERSION 3.0-1

OVERVIEW

    o Minor changes to placate the package checker.


        CHANGES IN spatstat.model VERSION 3.0-0

OVERVIEW

    o New package

    o We thank Achmad Choiruddin and Suman Rakshit for contributions.
    
    o Regularized model-fitting in 'ppm' and 'kppm'.
    
    o integral.msr accepts a weight function.

    o Weighted version of a measure.

    o Residuals for recursively-partitioned models.

    o Residuals for any estimate of intensity.

    o U-shaped curves in 'rhohat'.
    
    o Bug fixes and minor improvements.
    
NEW FUNCTIONS

    o measureWeighted
    Weighted version of a measure.

    o residuals.rppm
    Residual measure for a recursively-partitioned point process model.

    o residualMeasure
    Residual measure given an observed point pattern
    and an estimate of its intensity.

SIGNIFICANT USER-VISIBLE CHANGES

    o Package structure
    The package 'spatstat.core' has been split into two packages called
    'spatstat.explore' (for exploratory data analysis)
    and 'spatstat.model' (for modelling and formal inference).

    o spatstat.model
    The new package 'spatstat.model' contains the code for model-fitting,
    model diagnostics, and formal inference. Examples include 'ppm',
    'kppm', 'mppm', 'dppm', 'slrm', 'simulate.ppm', 'anova.ppm',
    'diagnose.ppm', 'residuals.ppm', 'leverage.ppm', 'addvar', 'parres', 

    o NEWS
    The NEWS file for the new package 'spatstat.model'
    contains older news items from the defunct package 'spatstat.core'
    (for functions which are now in 'spatstat.model').

    o ppm
    New argument 'improve.type'.

    o ppm
    Now supports regularized model-fitting when 'improve.type="enet"'.

    o ppm
    Option 'method="ho"' is replaced by 'improve.type="ho"'.

    o ppm
    Huang-Ogata approximate maximum likelihood can be applied to logistic fits
    by setting 'method="logi"' and 'improve.type="ho"'.
    
    o kppm
    New argument 'ppm.improve.type'.

    o kppm
    Now supports regularized model-fitting of the first order trend
    when 'ppm.improve.type="enet"'.

    o integral.msr
    New argument 'weight' specifies a weight (integrand) for the integration.

    o rhohat.ppm, rhohat.slrm
    New options 'smoother="mountain"' and 'smoother="valley"'
    for estimating a unimodal function (U-shaped curve).

    o rhohat.ppm, rhohat.slrm
    If the unit of length is a 'distfun', the name of the unit of length
    is saved and displayed on the plot.

    o rhohat.ppm, rhohat.slrm
    New arguments 'jitter', 'jitterfactor', 'interpolate' allow greater
    control over the calculation.
    
    o rhohat.ppm, rhohat.slrm
    New argument 'do.CI' specifies whether to calculate confidence bands.
    
BUG FIXES

    o predict.ppm
    Argument 'new.coef' was ignored in the calculation of the standard error
    when 'se=TRUE'.
    Fixed.
    
    o predict.ppm
    Argument 'new.coef' was ignored in calculating the standard error
    (and therefore the width of the interval) when 'type="count"'
    and 'interval="confidence"' or 'interval="prediction"'.
    Fixed.

    o vcov.mppm
    Crashed with a message about 'subscript out of bounds', for some models.
    Fixed.
        CHANGES IN spatstat.core VERSION 2.4-4.010

OVERVIEW

    o Internal improvements.

        CHANGES IN spatstat.core VERSION 2.4-4

OVERVIEW

    o Bug fixes and minor improvements.


        CHANGES IN spatstat.core VERSION 2.4-3

OVERVIEW

    o We thank Art Stock for contributions.
    
    o Bug fixes and minor improvements.

        CHANGES IN spatstat.core VERSION 2.4-2

OVERVIEW

    o Internal bug fixes.

        CHANGES IN spatstat.core VERSION 2.4-1

OVERVIEW

    o We thank Frederic Lavancier, Sebastian Meyer and Sven Wagner
    for contributions.
    
    o Improved approximation of intensity of Gibbs models.
    
    o Experimental code to represent (theoretical) point process models
    
    o Extract more information about a point process model.

    o Internal improvements and bug fixes.

NEW CLASSES

    o zgibbsmodel
    Experimental. An object of class 'zgibbsmodel' represents a
    Gibbs point process model with specified parameter values
    (whereas 'ppm' represents a model fitted to data).

NEW FUNCTIONS

    o hardcoredist
    Extract the hard core distance of a point process model.

    o interactionorder
    Extract the order of interpoint interaction of a point process model.

    o zgibbsmodel
    Experimental. Create an object of class 'zgibbsmodel'.

    o print.zgibbsmodel
    Experimental. Print an object of class 'zgibbsmodel'.
    
    o is.poisson.zgibbsmodel, is.stationary.zgibbsmodel
    Experimental. Methods for class 'zgibbsmodel'

SIGNIFICANT USER-VISIBLE CHANGES

    o intensity.ppm
    Can now calculate the Coeurjolly-Lavancier DPP approximation of intensity.
    [Code kindly contributed by Frederic Lavancier]
    New argument 'approx' specifies the choice of approximation.

BUG FIXES

    o vcov.ppm
    Crashed in some cases, with message 'object lamdel not found'.
    [Spotted by Sven Wagner.]
    Fixed.
    
        CHANGES IN spatstat.core VERSION 2.4-0

OVERVIEW

    o We thank Sriram Ramamurthy for contributions.

    o spatstat.core now depends on the new package 'spatstat.random'.

    o Functions for generating random patterns have been removed.
    
    o Important bug fixes in anova.mppm and vcov.mppm.

    o Minor improvements and bug fixes

SIGNIFICANT USER-VISIBLE CHANGES

    o package structure
    The code for generating random spatial patterns (including 'rpoispp',
    'rMatClust', 'rThomas', 'rNeymanScott', 'rStrauss', 'rmh')
    has been removed from 'spatstat.core' and placed in a new package
    'spatstat.random'. This new package is required by 'spatstat.core'.

    o anova.mppm
    Improved labelling of models in output.

    o qqplot.ppm, plot.qqppm
    Improved the text label indicating the type of residuals.

BUG FIXES

    o vcov.mppm
    For Gibbs (non-Poisson) models, the variance matrix was
    calculated incorrectly in some cases. 
    Fixed.

    o anova.mppm
    Results were sometimes incorrect if the two models
    had different interactions (e.g. Strauss vs Poisson).
    Fixed.

    o anova.mppm
    Crashed for some models with a message about 'coefficient missing
    from new.coef'.
    Fixed.
    
    o anova.mppm
    Gave a warning for some models about "Internal error: 
    unable to map submodels to full model".
    Fixed.

    o addvar
    If the covariate contained NA, NaN or Infinite values,
    the calculations were sometimes incorrect.
    Fixed.

    o pcfmodel.ppm
    Refused to handle an inhomogeneous Poisson process.
    Fixed.

    o fitted.ppm
    Crashed if leaveoneout=TRUE and the model had no fitted coefficients.
    Fixed.

        CHANGES IN spatstat.core VERSION 2.3-2

OVERVIEW

    o More diagnostics for spatial logistic regression models.

    o Important bug fix in kppm.

    o Increased numerical stability in kppm.
    
    o Minor improvements and bug fixes.

    o We thank Jonas Brehmer for contributions.

NEW FUNCTIONS

    o lurking.slrm
    Lurking variable plot for spatial logistic regression models.

    o eem.slrm
    Exponential energy marks for spatial logistic regression models.

    o eem.ppm
    Exponential energy marks for Gibbs and Poisson point process models
    (this function was previously called 'eem').

SIGNIFICANT USER-VISIBLE CHANGES

    o eem
    The function 'eem' is now generic, with methods for 'ppm' and 'slrm'.
    The function previously named 'eem' is now called 'eem.ppm'.

    o objsurf.dppm, objsurf.kppm, objsurf.mincontrast
    New arguments 'xlim', 'ylim' determine the range of parameter values
    to be considered.

    o Hybrid
    Printed output from hybrid models has been improved slightly.

    o kppm
    New default settings ensure greater numerical stability of the
    optimization algorithm against the effects of the scale of the
    spatial coordinates.  New argument 'stabilize' specifies whether
    the optimization algorithm should be numerically stabilized.

    o pcf
    Improved error message
    
BUG FIXES

    o kppm
    Results were sometimes incorrect for method='clik2' and method='palm'
    because the log composite likelihood was erroneously truncated to
    positive values. Any fitted model for which logLik(model) = 2.2e-16
    should be suspected of being incorrect.
    Fixed.

    o MultiHard
    A hybrid of 'MultiHard' with another multitype interaction caused an error.
    Fixed.

    o simulate.dppm, simulate.detpointprocfamily
    Crashed, rarely, with error message 'The dimension differs from
    the number of columns in index'.
    Fixed.
    
    o as.fv.kppm, as.fv.dppm
    The default plot labels in 'as.fv(x)' implied that the model 'x'
    was inhomogeneous.
    Fixed.
    
        CHANGES IN spatstat.core VERSION 2.3-1

OVERVIEW

    o Covariates in ppm and mppm may be functions that depend on the marks
    as well as the spatial coordinates.

    o Automatic selection of threshold for defining a binary predictor.
    
    o Random perturbation of line segments.
    
    o Minor extensions, performance improvements, and bug fixes.

NEW FUNCTIONS

    o coef<-.fii
    Changes the coefficients of a fitted interaction object
    (a method for the generic "coef<-")

SIGNIFICANT USER-VISIBLE CHANGES

    o ppm
    Covariates which are functions may now depend on the marks
    as well as the spatial coordinates: function(x,y,marks).
    
    o mppm
    Covariates which are functions may now depend on the marks
    as well as the spatial coordinates: function(x,y,marks).
    
    o reach.kppm
    Now returns a result for LGCP models as well as cluster models.

    o update.kppm
    New argument 'envir'.
    
    o clusterfit
    Minor changes to the argument list.

    o mincontrast
    Minor changes to the argument list.

    o mincontrast
    Improved algorithm for handling NA, NaN or infinite values.

    o varcount
    Argument B has a sensible default.

BUG FIXES

    o subfits
    The fitted interaction coefficients were garbled.
    If 'mfit' is the mppm object and 'a <- subfits(mfit)[[i]]' is one of the
    sub-models, then coef(a) was correct, but coef(fitin(a)) was incorrect.
    The fitted interaction was shown correctly by printing 'mfit'
    but incorrectly by printing 'a'.
    Fixed.
    
    o varcount
    The result was very inaccurate if the cluster radius was small 
    compared to the size of the window 'B', due to discretisation error.
    Fixed.

        CHANGES IN spatstat.core VERSION 2.3-0

OVERVIEW

    o We thank Chiara Fend for contributions.
    
    o Extensive support for spatial logistic regression models.

    o New fitting method in kppm and dppm.

    o Summary method for 'objsurf'.

    o Minor bug fixes.

NEW FUNCTIONS

    o response
    Generic function which extracts the values of the response
    in a fitted model. There are methods for lm, glm (which extract the
    numeric vector of responses), and ppm, kppm, slrm, lppm and mppm
    (which extract the original data point pattern).

    o summary.objsurf, print.summary.objsurf
    Summary method for 'objsurf'

    o residuals.slrm
    Residuals for spatial logistic regression models.
    
    o leverage.slrm, influence.slrm, dfbetas.slrm, dffit.slrm
    Leverage and influence diagnostics for spatial logistic regression models.
    
    o rhohat.slrm
    Method for 'rhohat' for spatial logistic regression models.

    o envelope.slrm
    Method for 'envelope' for spatial logistic regression models.

    o intensity.slrm
    Method for 'intensity' for spatial logistic regression models.

    o deviance.slrm
    Method for 'deviance' for spatial logistic regression models.

    o pseudoR2.slrm
    Method for 'pseudoR2' for spatial logistic regression models.

    o quadrat.test.slrm
    Method for 'quadrat.test' for spatial logistic regression models.

    o parameters.slrm
    Method for 'parameters' for spatial logistic regression models.

    o valid.slrm
    Method for 'valid' for spatial logistic regression models.

    o emend.slrm
    Method for 'emend' for spatial logistic regression models.

    o roc.slrm
    Method for 'roc' for spatial logistic regression models.

    o auc.slrm
    Method for 'auc' for spatial logistic regression models.

    o Window.slrm, as.owin.slrm
    Methods for 'Window' and 'as.owin' for spatial logistic regression models.

SIGNIFICANT USER-VISIBLE CHANGES

    o kppm, dppm
    New option 'method="adapcl"' performs adaptive composite likelihood fitting.
    [Contributed by Chiara Fend.]
    
BUG FIXES

    o clusterfield
    Values of the cluster field were slightly incorrect (slightly higher
    than the correct values) near the edge of the window,
    because an 'edge correction' was mistakenly applied.
    Fixed.

        CHANGES IN spatstat.core VERSION 2.2-0

OVERVIEW

    o summary method for spatial logistic regression models

NEW FUNCTIONS

    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.

        CHANGES IN spatstat.core VERSION 2.1-2

OVERVIEW

    o Reduced CRAN check time.

        CHANGES IN spatstat.core VERSION 2.1-1

OVERVIEW

    o Minor bug fix

BUG FIXES

    o simulate.kppm
    Conditional simulation crashed on rare occasions,
    with an error about negative probabilities.
    Fixed.

        CHANGES IN spatstat.core VERSION 2.1-0

OVERVIEW

    o We thank Achmad Choiruddin, Jean-Francois Coeurjolly
    and Rasmus Waagepetersen for contributions.

    o Conditional simulation in kppm
    
    o Information criteria for model selection in kppm

    o Modified handling of covariates in slrm

    o Improved output in summary.mppm

    o Minor improvements and bug fixes.

NEW FUNCTIONS

    o ic
    Information criteria for model selection in ppm and kppm.
    Kindly contributed by Achmad Choiruddin, Jean-Francois Coeurjolly
    and Rasmus Waagepetersen.
    
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 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 pairs.im
    New argument 'drop'.
    
BUG FIXES

    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.

        CHANGES IN spatstat.core 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 More support for spatial logistic regression models.

    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.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o slrm
    'step' can now be applied to models fitted using 'slrm'.

    o predict.mppm
    Now supports multitype point process models.

    o predict.mppm
    Improved handling of argument 'newdata'
    
    o simulate.ppm
    Now recognises the argument 'window' as an alternative to 'w'.

    o kppm
    Improved numerical robustness.

    o anova.mppm
    Issues a warning when applied to random-effects models
    (models fitted using the argument 'random').

    o mincontrast
    New argument 'action.bad.values' specifies what action is taken
    when the summary function produces NA or NaN or infinite values.
    

BUG FIXES

    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 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.
    
    
        CHANGES IN spatstat.core VERSION 1.65-11

OVERVIEW

    o Internal tweaks.
    
        CHANGES IN spatstat.core VERSION 1.65-10

OVERVIEW

    o Minor corrections to documentation.

        CHANGES IN spatstat.core VERSION 1.65-9

OVERVIEW

    o We thank Ian Buller for a suggestion.

        CHANGES IN spatstat.core VERSION 1.65-8

OVERVIEW

    o Minor changes to appease the compiler.

        CHANGES IN spatstat.core VERSION 1.65-7

OVERVIEW

    o We thank Michael Chirico for a contribution.

    o Minor changes to appease the compiler.
    
        CHANGES IN spatstat.core VERSION 1.65-6

OVERVIEW

    o We thank Tilman Davies and Pavel Fibich for contributions.

    o Important bug fix in simulation of log-Gaussian Cox processes.

    o Increased speed for large datasets.

    o variance calculations handle larger datasets.
    
SIGNIFICANT USER-VISIBLE CHANGES

    o vcov.ppm, summary.ppm
    Variance calculations now handle larger datasets
    (using sparse arrays).
    
    o overall speed
    Changes have been made to the internal code of spatstat 
    which should accelerate computations involving large datasets.

    o localpcf, localpcfinhom
    New argument 'rvalue'.

BUG FIXES

    o 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 print.ppm
    Crashed sometimes when applied to the result of subfits().
    Fixed.
    
    CHANGES IN spatstat.core VERSION 1.65-5

OVERVIEW

    o Minor changes required by CRAN.
    

        CHANGES IN spatstat.core VERSION 1.65-1

OVERVIEW

    o Added NEWS file.


        CHANGES IN spatstat.core VERSION 1.65-0

OVERVIEW

    o Package initialised at version 1.65-0

SIGNIFICANT USER-VISIBLE CHANGES

    o spatstat.core
    The package 'spatstat.core' has been created from a subset
    of the code in the original 'spatstat' package version 1.65-0.
    It contains the core functionality for statistical analysis of
    spatial data. For an overview, see help("spatstat.core-package")

    o Execution
    The 'spatstat.core' package is slightly faster than the corresponding
    code in the 'spatstat' package, because the procedure for calling
    internal C functions has been streamlined.