Description
Class for RandomField's representation of model estimation
results
Usage
"residuals"(object, ..., method="ml", full=FALSE)
"summary"(object, ..., method="ml")
"plot"(x, y, ...)
"contour"(x, ...)
"contour"(x, ...)
RFhessian(model)
Arguments
object
see the generic function;
...
-
plot: arguments to be passed to methods; mainly graphical
arguments, or further models in case of class 'RMmodel',
see Details.
-
summary: see the generic function
-
contour : see RFplotEmpVariogram
method
character; only for class(x)=="RFfit"; a
vector of slot names for which the fitted covariance or variogram
model is to be plotted; should be a subset of
slotNames(x) for which the corresponding slots are of class
"RMmodelFit"; by default, the maximum likelihood fit
("ml") will be
plotted
full
logical.
if TRUE submodels are reported as well (if available).
x
object of class RFsp or
RFempVario or
RFfit or
RMmodel; in the latter case, x can
be any sophisticated model but it must be either stationary or a
variogram model model
class(x)=="RF_fit" or class(x)=="RFfit", obtained
from RFfit
Creating Objects
Objects are created by the function
RFfitSlots
autostart:- RMmodelFit; contains the estimation results
for the method 'autostart' including a likelihood value, a constant
trend and the residuals
boxcox:- logical; whether the
parameter of a Box Cox tranformation has been estimated
coordunits:- string giving the units of the coordinates,
see also option
coordunits of RFoptions.
deleted:- integer vector.
positions of the parameters that has been deleted to get the set of
variables, used in the optimazation
ev:- list; list of objects of class
RFempVariog,
contains the empirical variogram estimates of the data fixed:-
list of two vectors. The fist gives the position where the
parameters are set to zero. The second gives the position where the
parameters are set to one.
internal1:- RMmodelFit; analog to slot 'autostart'
internal2:- RMmodelFit; analog to slot 'autostart'
internal3:- RMmodelFit; analog to slot 'autostart'
lowerbounds:- RMmodel; covariance model in which each
parameter value gives the lower bound for the respective parameter
ml:- RMmodelFit; analog to slot 'autostart'
modelinfo:- Table with information on the parameters:
name, boundaries, type of parameter
n.covariates:- number of covariates
n.param:-
number of parameters (given by the user)
n.variab:-
number of variables (used internally);
n.variab is always less than or equal to n.param
number.of.data:-
the number of data values passed to
RFfit that are
not NA or NaN
number.of.parameters:-
total number of parameters of the model that had to be estimated
including variances, scales, co-variables, etc.
p.proj:- vector of integers. The original position of those
parameters that are used in the submodel
plain:- RMmodelFit; analog to slot 'autostart'
report:-
if not empty, it indicates that this model should be reported
and gives a standard name of the model. Various function, e.g.
print.RMmodelFit uses
this information if their argument full equals TRUE. self:- RMmodelFit; analog to slot 'autostart'
sd.inv:- RMmodelFit; analog to slot 'autostart'
sqrt.nr:- RMmodelFit; analog to slot 'autostart'
submodels:-
list. Sequence (in some cases even nested sequence)
of models that is used to determine an initial value in
table:- matrix; summary of estimation results of
different methods
transform:- function;
true.tsdim:-
time space dimension of the (original!) data,
even for submodels that consider parts of separable models.
true.vdim:-
multivariability of the (original!) data,
even for submodels that consider independent models
for the multivariate components.
upperbounds:- RMmodel; see slot 'lowerbounds'
users.guess:- RMmodelFit; analog to slot 'autostart'
ml:- RMmodelFit; analog to slot 'autostart'; with maximum
likelihood method
v.proj:- vector of integers.
The components selected in one of the submodels
varunits:- string giving the units of the variables,
see also option
varunits of RFoptions.
x.proj:-
logical or integer. If logial, it means that no
separable model is considered there. If integer, then
it gives the considered directions of a separable model
Z:-
standardized list of information on the data
Methods
- plot
signature(x = "RFfit"): gives a plot of the
empirical variogram together with the fitted model, for more details see
plot-method.
- show
signature(x = "RFfit"): returns the structure
of x
- persp
- codesignature(obj =
"RFfit"): generates
persp plots
- print
signature(x = "RFfit"): identical with
show-method, additional argument is max.level
- [
signature(x = "RFfit"): enables accessing
the slots via the "["-operator, e.g. x["ml"]
- as
signature(x = "RFfit"):
converts into other formats, only implemented for target class
RFempVariog
- anova
- performs a likelihood ratio test base on a chisq approximation
- summary
- provides a summary
- logLik
- provides an object of class
"logLik"
- AIC,BIC
- provides the AIC and BIC information, respectively
signature(x = "RFfit", y = "missing")- Combines the plot of
the empirical variogram with the estimated covariance or variogram
model (theoretical) curves; further models can be added via the
argument
model.
Further 'methods'
AICc.RFfit(object, ..., method="ml", full=FALSE)] AICc.RF_fit(object, ..., method="ml", full=TRUE)References
AICc:
- Hurvich, C.M. and Tsai, C.-L. (1989)
Regression and Time Series Model Selection in Small Samples
Biometrika, 76, 297-307.