We generalize function returnlevelplot
from package distrMod to
be applicable to distribution and probability model objects. In this context,
returnlevelplot
produces a rescaled QQ plot of data (argument x
)
against a (model) distribution. For arguments y
of class RobModel
,
points at a high “distance” to the model
are plotted smaller. For arguments y
of class kStepEstimate
,
points at with low weight in the [p]IC are plotted bigger and their
color gets faded out slowly. This parallels the behaviour of the respective
qqplot
methods.
Graphical parameters may be given as arguments to returnlevelplot
.
returnlevelplot(x, y, ...)
# S4 method for ANY,RobModel
returnlevelplot(x, y,
n = length(x), withIdLine = TRUE, withConf = TRUE,
withConf.pw = withConf, withConf.sim = withConf,
plot.it = TRUE, xlab = deparse(substitute(x)),
ylab = deparse(substitute(y)), ..., distance = NormType(),
n.adj = TRUE)
# S4 method for ANY,InfRobModel
returnlevelplot(x, y, n = length(x), withIdLine = TRUE,
withConf = TRUE, withConf.pw = withConf, withConf.sim = withConf,
plot.it = TRUE, xlab = deparse(substitute(x)), ylab =
deparse(substitute(y)), ..., cex.pts.fun = NULL, n.adj = TRUE)
# S4 method for ANY,kStepEstimate
returnlevelplot(x, y,
n = length(x), withIdLine = TRUE, withConf = TRUE,
withConf.pw = withConf, withConf.sim = withConf,
plot.it = TRUE, xlab = deparse(substitute(x)),
ylab = deparse(substitute(y)), ...,
exp.cex2.lbs = -.15,
exp.cex2.pts = -.35,
exp.fadcol.lbs = 1.85,
exp.fadcol.pts = 1.85,
bg = "white")
As for function returnlevelplot
from package stats.
data to be checked for compatibility with distribution/model y
.
object of class "RobModel"
, of class "InfRobModel"
or of
class "kStepEstimate"
.
numeric; number of quantiles at which to do the comparison.
logical; shall line y = x
be plotted in?
logical; shall confidence lines be plotted?
logical; shall pointwise confidence lines be plotted?
logical; shall simultaneous confidence lines be plotted?
logical; shall be plotted at all (inherited from
returnlevelplot
)?
x-label
y-label
further parameters for method returnlevelplot
with signature
ANY,ProbFamily
(see returnlevelplot
) or with function
plot
rescaling function for the size of the points to be plotted;
either NULL
(default), then log(1+abs(x))
is used,
or a function which is then used.
logical; shall sample size be adjusted for possible outliers according to radius of the corresponding neighborhood?
a function mapping observations x
to the positive reals;
used to determine the size of the plotted points (the larger distance(x)
,
the smaller the points are plotted.
for objects kStepEstimate
based on a [p]IC of class HampIC
:
exponent for the weights of this [p]IC used to magnify the labels.
for objects kStepEstimate
based on a [p]IC of class HampIC
:
exponent for the weights of this [p]IC used to magnify the symbols.
for objects kStepEstimate
based on a [p]IC of class HampIC
:
exponent for the weights of this [p]IC used to find out-fading colors.
for objects kStepEstimate
based on a [p]IC of class HampIC
:
exponent for the weights of this [p]IC used to find out-fading colors.
background color to fade against
Peter Ruckdeschel peter.ruckdeschel@uni-oldenburg.de
signature(x = "ANY", y = "RobModel")
:
produces a QQ plot of a dataset x
against the theoretical
quantiles of distribution of robust model y
.
signature(x = "ANY", y = "InfRobModel")
:
produces a QQ plot of a dataset x
against the theoretical
quantiles of distribution of infinitesimally robust model y
.
signature(x = "ANY", y = "kStepEstimate")
:
produces a QQ plot of a dataset x
against the theoretical
quantiles of the model distribution of model at which
the corresponding kStepEstimate
y
had been calibrated at.
By default, if the [p]IC of the kStepEstimate
is of class
HampIC
, i.e.; has a corresponding weight function,
points (and, if withLab==TRUE
, labels) are
scaled and faded according to this weight function. Corresponding
arguments exp.cex2.pts
and exp.fadcol.pts
control this
scaling and fading, respectively
(and analogously exp.cex2.lbs
and exp.fadcol.lbs
for the labels).
The choice of these arguments has to be done on a case-by-case basis.
Positive exponents induce fading, magnification with increasing weight,
for negative exponents the same is true for decreasing weight; higher
(absolute) values increase the speed of fading / magnification.
ismev: An Introduction to Statistical Modeling of Extreme Values. R package version 1.39. https://CRAN.R-project.org/package=ismev; original S functions written by Janet E. Heffernan with R port and R documentation provided by Alec G. Stephenson. (2012).
Coles, S. (2001). An introduction to statistical modeling of extreme values. London: Springer.
qqplot
from package stats -- the standard QQ plot
function, returnlevelplot
from package distrMod (which
is called intermediately by this method), as well as
qqbounds
, used by returnlevelplot
to produce confidence
intervals.
returnlevelplot(rnorm(40, mean = 15, sd = sqrt(30)), Chisq(df=15))
RobM <- InfRobModel(center = NormLocationFamily(mean=13,sd=sqrt(28)),
neighbor = ContNeighborhood(radius = 0.4))
# \donttest{
## \donttest to reduce check time
x <- rnorm(20, mean = 15, sd = sqrt(30))
returnlevelplot(x, RobM)
returnlevelplot(x, RobM, alpha.CI=0.9, add.points.CI=FALSE)
# }
## further examples for ANY,kStepEstimator-method
## in example to roptest() in package ROptEst
Run the code above in your browser using DataLab