Documentation of major changings

Description

This man page documents some major changings in RandomFields.

Changes done in 3.1.0 (Summer 2015)

• full (univariate) trend modelling

• error in particular in RFfit corrected

• RFfit runs much faster now

• effects of modus operandi changed for estimating

Corrections done in 3.0.56 (Jan 2015)

• log Gauss field corrected (has been a log log Gauss field)

• RMconstant is now called RMfixcov

Corrections done in 3.0.55 (Jan 2015)

• Conditional simulation: several severe typos corrected.

Major Revision: changings from Version 2 to Version 3 (Jan 2014)

• S4 objects

  • RandomFields is now based on S4 objects using the package sp. The functions accept both sp objects and simple objects as used in version 2. See also above.

• Documentation

  • each model has now its own man page;

  • classes of models and functions are bundled in several pages: Covariance models start with RM, distribution families with RR, processes with RP, user functions with RF

  • the man pages of several functions are split into two parts:

    (i) a beginners man page which includes a link to

    (ii) man pages for advanced users

• Interfaces

  • The interfaces become simpler, at the same time more powerful than the functions in version 2. E.g., RFsimulate can perform unconditional simulation, conditional simulation and random imputing.

  • Only those arguments are kept in the functions that are considered as being absolutely necessary. All the other arguments can be included as options.

  • RFgui is an instructive interface based on tcl/tk, replacing the former ShowModels

• Inference for Gaussian random fields

  • RFfit has undergone a major revision. E.g.:

    (i) estimation of random effect models with spatial covariance structure

    (ii) automatic estimation of 10 and more arguments in multivariate and/or space-time models

  • RFvariogram is now based on an fft algorithm if the data are on a grid, even allowing for missing values.

  • RFratiotest has been added.

• Processes

  • Modelling of maxstable processes has been enhanced, including

    (i) the simulation of Brown-Resnick processes

    (ii) initial support of tail correlation functions;

  • Further processes chi2 processes, compound Poisson processes, binary processes added.

• Models

  • the formula notation for linear models may now be defined

  • Novel, user friendly definition of the covariance models

  • Multivariate and vector-valued random fields are now fully included

  • The user may now define his own functions, to some extend.

  • The trend allows for much more flexibility

  • Distributions may now be included which will be extended to Bayesian modelling in future.

Author(s)

Martin Schlather, schlather@math.uni-mannheim.de, https://www.wim.uni-mannheim.de/schlather/

Examples

RFoptions(seed=0) ## *ANY* simulation will have the random seed 0; set
##                   RFoptions(seed=NA) to make them all random again

## S4 vs S3
x <- seq(0, 10, 0.1)
model <- RMexp()
plot(RFsimulate(model, x)) ## S4
plot(RFsimulate(model, x, spConform=FALSE)) ## no class