Modelling Spatial Extremes

Tools for the statistical modelling of spatial extremes using max-stable processes, copula or Bayesian hierarchical models. More precisely, this package allows (conditional) simulations from various parametric max-stable models, analysis of the extremal spatial dependence, the fitting of such processes using composite likelihoods or least square (simple max-stable processes only), model checking and selection and prediction. Other approaches (although not completely in agreement with the extreme value theory) are available such as the use of (spatial) copula and Bayesian hierarchical models assuming the so-called conditional assumptions. The latter approaches is handled through an (efficient) Gibbs sampler. Some key references: Davison et al. (2012) , Padoan et al. (2010) , Dombry et al. (2013) .


News

	CHANGES IN SpatialExtremes VERSION 2.0-7

BUG FIXES

o Fix a bug in the (exact!) simulation from the Schlather max-stable process.

o Fix a bug in how to handle NA values in the prediction



	CHANGES IN SpatialExtremes VERSION 2.0-6

BUG FIXES

o Minor modification to suppress warnings with a new release of R.


	CHANGES IN SpatialExtremes VERSION 2.0-5

BUG FIXES

o Fix a bug with the simulation for the geometric model. Thanks to Zhuo Wang for finding it!


	CHANGES IN SpatialExtremes VERSION 2.0-4

NEW FEATURES

o Exact simulation for max-stable processes are now fully implemented.

o Parallelization is used for computing pairwise likelihood---should be faster (finally).


	CHANGES IN SpatialExtremes VERSION 2.0-3

NEW FEATURES

o Concurrence probabilities and cells are now implemented.

o Exact simulation for max-stable processes are (partially) implemented.


	CHANGES IN SpatialExtremes VERSION 2.0-0

SIGNIFICANT USER-VISIBLE CHANGES

o There is only one methodology to compute standard errors; hence
the arguments std.err.type is deprectated.

NEW FEATURES

o Missings values (NA's) are now available for max-stable and latent models !

BUG FIXES

o Fix a bug with the mean of the prior distribution assumed on the
regression parameters for the latent variable approach.


	CHANGES IN SpatialExtremes VERSION 1.9-0

SIGNIFICANT USER-VISIBLE CHANGES

o The memory allocation is now much improved, your computer should
not freeze anymore ;-)

NEW FEATURES

o Conditional simulations from various max-stable models (not
max-linear models) are available.

o Simulation from the extremal-t process are now performed using
its spectral characterization

o Different (but still naive) approaches for simulating
Brown-Resnick processes are implemented

BUG FIXES

o Fix a bug with the latent variable model that might crash R

o Fix a bug with temporal trends



	CHANGES IN SpatialExtremes VERSION 1.8-1

SIGNIFICANT USER-VISIBLE CHANGES

o The parameter "sill" is now called "nugget" to avoid
confusion. More precisely, we now have nugget = 1 - sill, so if
one wants no nugget effect just set nugget = 0.

NEW FEATURES

o Model checking for the copula based models is now implemented
--- similar to the plot.maxstab function

o It is now possible to simulate from copula based models

o The package now uses openmp as much as possible!

BUG FIXES

o Fix a bug with the Student-t copula model

o Fix a bug with standard errors and 3d Smith model


	CHANGES IN SpatialExtremes VERSION 1.8-0

SIGNIFICANT USER-VISIBLE CHANGES

o Improvement of the simulation technique for Brown-Resnick
processes.

NEW FEATURES

o (Approximate) conditional simulations of unit Frechet max-stable
random fields is implemented.

o The copula approach is now available --- use with caution!

o Max-linear models are now available.

BUG FIXES

o Fix a bug with model selection.



	CHANGES IN SpatialExtremes VERSION 1.7-0

SIGNIFICANT USER-VISIBLE CHANGES

o Update contact info: now in Montpellier!

o A new function ''latent'' for Bayesian hierarchical models

BUG FIXES

o Fix a bug in simulating Brown-Resnick processes



	CHANGES IN SpatialExtremes VERSION 1.6-0

SIGNIFICANT USER-VISIBLE CHANGES

o The extremal t model is now available --- including
simulation. Simulations on a grid might be really long though.

o Model checking for fitted max-stable processes is easier as a
method for plotting object of class "maxstab" is implemented.

o "fitspatgev" and "fitmaxstab" now allow for "temporal"
covariates, for example a linear trend can be put for the location
parameter.

NEW FEATURES

o Simulation of Brown-Resnick processes are available but should
*NOT* be considered as stable!

o Computation of the empirical variogram is implemented

o The fitting and simulation of max-stable processes is now much faster.

o The circulant embedding method is now implemented.

BUG FIXES

o Fix a bug when using temporal covariates.

o Fix a bug in simulating Gaussian random fields having a powered
exponential covariance function through the turning bands method.

o The parametrization of Brown-Resnick processes was not
consistent throughout the code. Now fixed.


	CHANGES IN SpatialExtremes VERSION 1.5-1

BUG FIXES

o Fix a bug in computing the standard errors for the geometric
gaussian model (marginal part)


	CHANGES IN SpatialExtremes VERSION 1.5-0

SIGNIFICANT USER-VISIBLE CHANGES

o	The 'map' and 'condmap' functions now allow covariates

o	fitmaxstab allows weights in the pairwise likelihood

o	predict will display standard errors if possible

NEW FEATURES

o	Brown-Resnick processes are available

o	The generalized Cauchy correlation function is available

o	New function 'lsmaxstab' that wraps 'fitcovariance' and
'fitcovmat' and is similar to 'fitmaxstab' but uses least squares

o	Composite version of the BIC is available

o	The standard errors are now computed without the use of the
finite difference based estimate of the Hessian matrix

DEPRECATED & DEFUNCT

o The 'fitcovmat' and 'fitcovariance' will be removed soon as the new 'lsmaxstab' wraps them

BUG FIXES

o	Fix a bug in simulating smith's max-stable process on a grid

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("SpatialExtremes")

2.0-9 by Mathieu Ribatet, 8 months ago


http://spatialextremes.r-forge.r-project.org/


Browse source code at https://github.com/cran/SpatialExtremes


Authors: Mathieu Ribatet [aut, cre] , Richard Singleton [ctb] , R Core team [ctb]


Documentation:   PDF Manual  


Task views: Analysis of Spatial Data


GPL (>= 2) license


Imports maps, fields


Imported by mobirep.

Suggested by eva.


See at CRAN