Max-stable processes are the extension of the extreme value theory to random fields. Consequently, they are good candidate to the analysis of spatial extremes. The strategy used in this package is to fit max-stable processes to data using composite likelihood.
In the future, the package will allow for non-stationarity as well as other approaches to model spatial extremes; namely latent variable and copula based approaches.
A package vignette has been writen to help new users. It can be
viewed, from the R console, by invoking
rgp: simulates gaussian random fields,rmaxstab: simulates max-stable random fields,fitspatgev: fits a spatial GEV model to data,fitmaxstab,lsmaxstab: fits
max-stable processes to data,latent: draws a Markov chain from a Bayesian
hierarchical model for spatial extremes,predict: allows predictions
for fitted max-stable processes,map,condmap: plot a map for GEV
parameter as well as return levels - or conditional return levelsanova,TIC,DIC: help users in model selection,madogram,fmadogram,lmadogram: are (kind of) variograms devoted to extremes,fitextcoeff: estimates semi-parametrically the
extremal coefficient,extcoeff: plots the evolution of the extremal
coefficient from a fitted max-stable process,rbpspline: fits a penalized spline with radial
basis function,gev2frech,frech2gev: transform
GEV (resp. Frechet) observation to unit Frechet (resp. GEV) onesgevmle,gpdmle: fit the GEV/GPD
distributions to data,distance: computes the distance between each
pair of locations,profile,profile2d: computes the profile
composite likelihood,covariance,variogram: computes
the covariance/semivariogram function.