R: Interactive Choice of Models and Parameters

ShowModels {RandomFields}

R Documentation

Interactive Choice of Models and Parameters

Description

ShowModels is an interactive plot for the selection of models and their one- or two-dimensional simulations; it also allows for the fitting of variogram models by eye.

Usage

ShowModels(x, y=NULL,
           covx=ifelse(is.null(empirical), diff(range(x))/5,
                       max(empirical$c)), 
           fixed.rs=TRUE, method=NULL, empirical=NULL,
           model=NULL, param=NULL, all.param=NULL,  
           legends = TRUE, 
           register=0, Mean=NULL, erase=TRUE,
           x.fraction=0.60, cex.names=1, covx.default = 100,
           link.fct=NULL, Zlim=NULL,
           Col.rect="red", Col.bg="blue", Col.sep="grey",
           Col.line="red", Col.txt="black", Col.flash="red",
           Col.vario="blue", Col.main="black",
           Col.model=c("red", "black"),  vario.lty=c(1,2), 
           cex.leg = 0.7 * cex.names,  cex.eval = 0.8 * cex.names,
           update=TRUE, screen.new=TRUE, use.outer.RFparameters=FALSE,
           debug=FALSE, ...)

Arguments

`x`	if `NULL` simulations are not performed; otherwise it gives the x coordinates of a grid as a sequence of increasing numbers
`y`	if `NULL` at most one-dimensional simulations are performed (depending on the value of `x`); otherwise `y` gives the y coordinates of a two-dimensional grid (as a sequence of increasing numbers).
`covx`	if a single value is given, it is the largest distance for which the covariance functions or the variograms are plotted; otherwise the models are plotted for the given values, and the origin.
`fixed.rs`	if `TRUE` then the same random seed is used for all simulations until the user clicks on the formula, the title or the subtitles.
`method`	simulation method, see RFMethods; if `NULL` then a suitable simulation method is chosen automatically.
`empirical`	empirical variogram; a list as returned by `EmpiricalVariogram`. Also empirical variograms with a pair number of anisotropy directions may be passed. Then the first and the middle one are taken.
`model`	covariance model, see `CovarianceFct`, or type `PrintModelList()` to get all options. If given, this model is shown at the beginning. Additive or multiplicative models are not allowed. However, model can also be given by a simple list definition, see `CovarianceFct`. Then `param` must not be given. In this case also anisotropic models can be defined.
`param`	parameter vector: `param=c(mean, variance, nugget, scale,...)`; the parameters must be given in this order; see `CovarianceFct` for more details. Only considered if `model` is given. If given, `model` is initialised by `param`.
`all.param`	`all.param=c(mean, variance, nugget, scale)`; the parameters must be given in this order; If `all.param` is given then the parameters of all covariance functions are set to the given values. The values are overwritten for a specific model if `model` and `param` are given. Note that it is not possible to set the values of additional (form) parameters of a parametrised class by means of `all.param`. In case of an anisotropic model the anisotropy matrix is by default diagonal with both entries equal to 1/`all.param[4]`.
`legends`	if `TRUE` then a legend is added to the two-dimensional plot.
`register`	register where intermediate results of the simulations are stored, see also `GaussRF`.
`Mean`	mean of the random field
`erase`	parameter of `split.screen`, which is called at the very beginning
`x.fraction`	the current screen is split into 2 x 2 screens. The parameter `x.fraction` gives the size of the left screens in the x directions as part of 1. See also the Details.
`cex.names`	font size for model names
`covx.default`	if `length(cov.x)==1` then [0, `cov.x`] is covered by `covx.default` points of equal distance
`link.fct`	`NULL` or `function(values)` or "MaxStable". Transformation of the Gaussian random field. If `link.fct="MaxStable"` then max-stable random fields are simulated for the given covariance function and the extremal coefficient function is given (up the constant -1) instead of the variogram or the covariance function
`Zlim`	Vector of two elements or list of two vectors of two elements. Graphical limits for the Gaussian random process (and the transformed field).
`Col.rect`	colour for interactive plot; see `eval.parameters`.
`Col.bg`	colour for interactive plot; see `eval.parameters`.
`Col.sep`	colour for interactive plot; see `eval.parameters`.
`Col.line`	colour for interactive plot; see `eval.parameters`.
`Col.txt`	colour for interactive plot; see `eval.parameters`.
`Col.flash`	colour for the previously chosen model
`Col.vario`	colour for the empirical variogram plot
`Col.main`	colour for the title of the random field
`Col.model`	vector of two colours for plotting the variogram of the Gaussian random field and the transformed field
`vario.lty`	vector of two line types for primary and secondary axis of the variogram
`cex.leg`	font size used in the legends
`cex.eval`	font size used in the menue entries
`update`	logical. If `TRUE` the plots are updated after each interactive change of the values. Otherwise, the bottom 'simulate' is added in the menu.
`screen.new`	logical. If `FALSE` the screen is erased before a simulation and completely rebuild; otherwise the screen is updated. If `FALSE` flickering appears during the update of the current screen, otherwise it may happen during the reorganisation of any window (and may take quite a lot of time).
`use.outer.RFparameters`	logical. If `FALSE` the following parameters usually set by `RFparameters` are internally set `PracticalRange=FALSE` `PrintLevel=1` if `debug=FALSE` and `5` otherwise. `maxstable.maxGauss=2` `CE.force=TRUE` `CE.trials=1` `CE.mmin=-4` `CE.useprimes=TRUE`
`debug`	logical. If `TRUE` then internally the `RFparameter()$PrintLevel` is set to 5.
`...`	additional graphics options for the plot of the one- or two-dimensional simulations, see `plot` and `image`.

Details

The interactive plot consists of 3 parts:

top left: graph of the covariance function or the variogram. In case empirical is given the empirical variogram is also plotted. If link.fct is given, then also the covariance function or the variogram is plotted. If the correlation model is for a non-stationary random field, the variogram for the transformed random field is not estimated in a primitive way – this is indicated with a star in the legend
bottom left: one- or two-dimensional simulation
right:

– list of implemented models; a specific model is chosen by the left mouse button, or:

– menu for the parameters of the chosen model. The list includes the variance, a nugget effect, the mean and the scale or the anisotropy parameters. Further, some global parameters can be changed. They are the PracticalRange (see RFparameters for details) and the angle of the main variogram direction (or NA, then it follows the angle of the anisotropy). Finally, the user can choose between the plot of the covariance and the corresponding variogram.

The interactive plot is left by clicking any mouse button different from the left when the top right part is active.

Value

list of the last model and its parameters.

Author(s)

Martin Schlather, martin.schlather@math.uni-goettingen.de http://www.stochastik.math.uni-goettingen.de/institute

Examples

# first example: one-dimensional simulations

options(locatorBell=FALSE)
x <-  seq(1,10,0.1);
ShowModels(x=x)

# second example: two-dimensional simulations and
#                 empirical variogram
dx <- runif(300,0,8)
dy <- runif(300,0,8)
dz <- GaussRF(x=dx, y=dy, grid=FALSE, model="gaus",
      param=c(1,2,1,2))
ev <- EmpiricalVariogram(x=dx, y=dy, data=dz, grid=FALSE,
      bin=(-1:20)/4)
x <-  seq(1,5,0.1);
ShowModels(x=x, y=x, empirical=ev)

# third example: two-dimensional anistropic simulations and
#                link function
x <- seq(1,10,0.1)
ShowModels(x=x, y=x, link=function(x) exp(x),
          model=list(list(model="spheric", var=1, aniso=c(1,0,0,5))))

x <-  seq(1,10,0.1)
ShowModels(x=x, link=function(x) exp(x),
           model=list(list(model="spheric",var=1, scale=1)))

x <-  seq(1,10,0.1)
ShowModels(x=x, link="MaxStable", fixed.rs=TRUE,
           model=list(list(model="gauss",var=1, scale=1)), type="l")

[Package RandomFields version 1.3.41 Index]