https://github.com/cran/RandomFields
Tip revision: e994a4415e67fa60cbfd3f208aaab20872521c0b authored by Martin Schlather on 14 February 2019, 21:02:19 UTC
version 3.3
version 3.3
Tip revision: e994a44
RFboxcox.Rd
\name{RFboxcox}
\alias{RFboxcox}
\title{Linear part of \command{\link{RMmodel}}}
\description{
\command{\link{RFboxcox}} performs the Box-Cox transformation:
\eqn{\frac{(x+\mu)^\lambda-1}{\lambda}}
}
\usage{
RFboxcox(data, boxcox, vdim = 1, inverse=FALSE, ignore.na=FALSE)
}
\arguments{
\item{data}{matrix or list of matrices.
}
\item{boxcox}{the one or two parameters \eqn{(\lambda, \mu)}
of the box cox transformation,
in the univariate case; if \eqn{\mu} is not given, then \eqn{\mu} is
set to \eqn{0}.
If not given, the globally defined parameters are used, see Details.
In the \eqn{m}-variate case \code{boxcox} should be a \eqn{2 \times
m} matrix. If \eqn{\lambda =\infty} then no transformation is performed.
}
\item{vdim}{the multivariate dimensionality of the field;
}
\item{inverse}{logical. Whether the inverse transformation should be performed.
}
\item{ignore.na}{logical. If \code{FALSE} an error message is returned
if any value of \code{boxcox} is \code{NA}. Otherwise the data are
returned without being transformed.
}
}
\details{
The Box-Cox transfomation \code{boxcox} can be set
globally through \command{\link{RFoptions}}. If it is set globally the
transformation applies in the \bold{Gaussian} case to
\command{\link{RFfit}},
\command{\link{RFsimulate}},
\command{\link{RFinterpolate}},
\command{\link{RFvariogram}}.
Always first, the Box-Cox transformation is applied to the data.
Then the command is performed. The result is back-transformed before
returned.
If the first value of the transformation is \code{Inf} no
transformation is performed (and is identical to \code{boxcox = c(1,0)}).
If \code{boxcox} has length 1, then the transformation parameter
\eqn{\mu} is set to \eqn{0}, which is the standard case.
}
\value{
\command{\link{RFboxcox}} returns a list
of three components, \code{Y}, \code{X}, \code{vdim} returning
the deterministic trend, the design matrix, and the multivariability,
respectively.
If \code{set} is positive, \code{Y} and \code{X} contain
the values for the \code{set}-th set of coordinates.
Else, \code{Y} and \code{X} are both lists containing
the values for all the sets.
}
\me
\seealso{
\link{Bayesian},
\command{\link{RMmodel}},
\command{\link{RFsimulate}},
\command{\link{RFlikelihood}}.
}
\references{
For the likelihood correction see
\itemize{
\item Konishi, S., and Kitagawa, G. (2008)
\emph{Information criteria and statistical modeling.}
Springer Science & Business Media. Section 4.9.
}
}
\examples{\dontshow{StartExample()}
RFoptions(seed=0) ## *ANY* simulation will have the random seed 0; set
## RFoptions(seed=NA) to make them all random again
data(soil)
str(soil)
soil <- RFspatialPointsDataFrame(
coords = soil[ , c("x.coord", "y.coord")],
data = soil[ , c("moisture", "NO3.N", "Total.N", "NH4.N", "DOC", "N20N")],
RFparams=list(vdim=6, n=1)
)
dta <- soil["moisture"]
\dontshow{if (RFoptions()$internal$examples_red) {
warning("data have been reduced !")
All <- 1:7
rm(soil)
data(soil)
soil <- RFspatialPointsDataFrame(
coords = soil[All, c("x.coord", "y.coord")],
data = soil[All, c("moisture", "NO3.N", "Total.N",
"NH4.N", "DOC", "N20N")],
RFparams=list(vdim=6, n=1)
)
dta <- soil["moisture"]
}}
model <- ~1 + RMplus(RMwhittle(scale=NA, var=NA, nu=NA), RMnugget(var=NA))
\dontshow{\dontrun{
## Assuming log-Gaussian Data
print(fit <- RFfit(model, data=dta, loggaus=TRUE))
}}
## main Parameter in the Box Cox transformation to be estimated
print(fit <- RFfit(model, data=dta, boxcox=NA))
\dontshow{FinalizeExample()}}
\keyword{spatial}