[Robast-commits] r1243 - pkg/ROptEst/man
noreply at r-forge.r-project.org
noreply at r-forge.r-project.org
Wed Nov 16 19:42:46 CET 2022
Author: stamats
Date: 2022-11-16 19:42:46 +0100 (Wed, 16 Nov 2022)
New Revision: 1243
Modified:
pkg/ROptEst/man/RMXEOMSEMBREOBRE.Rd
pkg/ROptEst/man/getIneffDiff.Rd
pkg/ROptEst/man/getMaxIneff.Rd
pkg/ROptEst/man/leastFavorableRadius.Rd
pkg/ROptEst/man/optIC.Rd
pkg/ROptEst/man/radiusMinimaxIC.Rd
pkg/ROptEst/man/roptest.Rd
Log:
changed an invalid URL detected by the CRAN checks during submission
Modified: pkg/ROptEst/man/RMXEOMSEMBREOBRE.Rd
===================================================================
--- pkg/ROptEst/man/RMXEOMSEMBREOBRE.Rd 2022-11-14 19:50:43 UTC (rev 1242)
+++ pkg/ROptEst/man/RMXEOMSEMBREOBRE.Rd 2022-11-16 18:42:46 UTC (rev 1243)
@@ -1,322 +1,322 @@
-\name{RMXEOMSEMBREOBRE}
-\alias{RMXEstimator}
-\alias{OMSEstimator}
-\alias{OBREstimator}
-\alias{MBREstimator}
-\title{ Optimally robust estimation: RMXE, OMSE, MBRE, and OBRE }
-\description{
- These are wrapper functions to 'roptest' to compute
- optimally robust estimates, more specifically RMXEs, OMSEs, MBREs, and OBREs,
- for L2-differentiable parametric families via k-step construction.
-}
-\usage{
-RMXEstimator(x, L2Fam, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
- steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
- OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
- withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
- IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
- withICList = getRobAStBaseOption("withICList"),
- withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
- initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
- withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
- withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
- diagnostic = FALSE)
-OMSEstimator(x, L2Fam, eps=0.5, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
- steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
- OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
- withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
- IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
- withICList = getRobAStBaseOption("withICList"),
- withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
- initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
- withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
- withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
- diagnostic = FALSE)
-OBREstimator(x, L2Fam, eff=0.95, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
- steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
- OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
- withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
- IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
- withICList = getRobAStBaseOption("withICList"),
- withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
- initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
- withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
- withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
- diagnostic = FALSE)
-MBREstimator(x, L2Fam, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
- steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
- OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
- withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
- IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
- withICList = getRobAStBaseOption("withICList"),
- withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
- initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
- withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
- withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
- diagnostic = FALSE)
-}
-\arguments{
- \item{x}{ sample }
- \item{L2Fam}{ object of class \code{"L2ParamFamily"} }
- \item{eff}{ positive real (0 <= \code{eff} <= 1): amount of asymptotic
- efficiency loss in the ideal model. See details below. }
- \item{eps}{ positive real (0 < \code{eps} <= 0.5): amount of gross errors.
- See details below. }
- \item{fsCor}{ positive real: factor used to correct the neighborhood radius;
- see details. }
- \item{initial.est}{ initial estimate for unknown parameter. If missing
- minimum distance estimator is computed. }
- \item{neighbor}{ object of class \code{"UncondNeighborhood"} }
- \item{steps}{ positive integer: number of steps used for k-steps construction }
- \item{distance}{ distance function used in \code{MDEstimator}, which in turn
- is used as (default) starting estimator. }
- \item{startPar}{ initial information used by \code{optimize} resp. \code{optim};
- i.e; if (total) parameter is of length 1, \code{startPar} is
- a search interval, else it is an initial parameter value; if \code{NULL}
- slot \code{startPar} of \code{ParamFamily} is used to produce it;
- in the multivariate case, \code{startPar} may also be of class \code{Estimate},
- in which case slot \code{untransformed.estimate} is used.}
- \item{verbose}{ logical: if \code{TRUE}, some messages are printed }
- \item{useLast}{ which parameter estimate (initial estimate or
- k-step estimate) shall be used to fill the slots \code{pIC},
- \code{asvar} and \code{asbias} of the return value. }
- \item{OptOrIter}{character; which method to be used for determining Lagrange
- multipliers \code{A} and \code{a}: if (partially) matched to \code{"optimize"},
- \code{getLagrangeMultByOptim} is used; otherwise: by default, or if matched to
- \code{"iterate"} or to \code{"doubleiterate"},
- \code{getLagrangeMultByIter} is used. More specifically,
- when using \code{getLagrangeMultByIter}, and if argument \code{risk} is of
- class \code{"asGRisk"}, by default and if matched to \code{"iterate"}
- we use only one (inner) iteration, if matched to \code{"doubleiterate"}
- we use up to \code{Maxiter} (inner) iterations.}
- \item{withUpdateInKer}{if there is a non-trivial trafo in the model with matrix \eqn{D}, shall
- the parameter be updated on \eqn{{\rm ker}(D)}{ker(D)}?}
- \item{IC.UpdateInKer}{if there is a non-trivial trafo in the model with matrix \eqn{D},
- the IC to be used for this; if \code{NULL} the result of \code{getboundedIC(L2Fam,D)} is taken;
- this IC will then be projected onto \eqn{{\rm ker}(D)}{ker(D)}.}
- \item{withPICList}{logical: shall slot \code{pICList} of return value
- be filled?}
- \item{withICList}{logical: shall slot \code{ICList} of return value
- be filled?}
- \item{na.rm}{logical: if \code{TRUE}, the estimator is evaluated at \code{complete.cases(x)}.}
- \item{initial.est.ArgList}{a list of arguments to be given to argument \code{start} if the latter
- is a function; this list by default already starts with two unnamed items,
- the sample \code{x}, and the model \code{L2Fam}.}
- \item{\dots}{ further arguments }
- \item{withLogScale}{logical; shall a scale component (if existing and found
- with name \code{scalename}) be computed on log-scale and backtransformed
- afterwards? This avoids crossing 0.}
- \item{..withCheck}{logical: if \code{TRUE}, debugging info is issued.}
- \item{withTimings}{logical: if \code{TRUE}, separate (and aggregate)
- timings for the three steps evaluating the starting value, finding
- the starting influence curve, and evaluating the k-step estimator is
- issued.}
- \item{withMDE}{ logical or \code{NULL}: Shall a minimum distance estimator be used as
- starting estimator---in addition to the function given in slot
- \code{startPar} of the L2 family? If \code{NULL} (default), the content
- of slot \code{.withMDE} in the L2 family is used instead to take
- this decision.}
- \item{withEvalAsVar}{logical or \code{NULL}: if \code{TRUE} (default), tells R
- to evaluate the asymptotic variance or if \code{FALSE} just to produces a call
- to do so. If \code{withEvalAsVar} is \code{NULL} (default), the content
- of slot \code{.withEvalAsVar} in the L2 family is used instead to take
- this decision.}
- \item{withMakeIC}{logical; if \code{TRUE} the [p]IC is passed through
- \code{makeIC} before return.}
- \item{modifyICwarn}{logical: should a (warning) information be added if
- \code{modifyIC} is applied and hence some optimality information could
- no longer be valid? Defaults to \code{NULL} in which case this value
- is taken from \code{RobAStBaseOptions}.}
- \item{E.argList}{\code{NULL} (default) or a list of arguments to be passed
- to calls to \code{E} from (a) \code{MDEstimator}
- (here this additional argument is only used if
- \code{initial.est} is missing), (b) \code{getStartIC},
- and (c) \code{kStepEstimator}. Potential clashes with
- arguments of the same name in \code{\dots} are resolved by inserting
- the items of argument list \code{E.argList} as named items, so
- in case of collisions the item of \code{E.argList} overwrites the
- existing one from \code{\dots}.}
- \item{diagnostic}{ logical; if \code{TRUE},
- diagnostic information on the performed integrations is gathered and
- shipped out as an attribute \code{diagnostic} of the return value
- of the estimators. }
-}
-\details{
- The functions compute optimally robust estimator for a given L2 differentiable
- parametric family; more specifically they are RMXEs, OMSEs, MBREs, and OBREs.
- The computation uses a k-step construction with an
- appropriate initial estimate; cf. also \code{\link[RobAStBase]{kStepEstimator}}.
- Valid candidates are e.g. Kolmogorov(-Smirnov) or von Mises minimum
- distance estimators (default); cf. Rieder (1994) and Kohl (2005).
-
- For OMSE, i.e., the asymptotically linear estimator with minimax mean squared
- error on this neighborhood of given size, the amount of gross errors
- (contamination) is assumed to be known, and is specified by \code{eps}.
- The radius of the corresponding infinitesimal
- contamination neighborhood is obtained by multiplying \code{eps}
- by the square root of the sample size.
-
- If the amount of gross errors (contamination) is unknown, RMXE should be used,
- i.e., the radius-minimax estimator in the sense of
- Rieder et al. (2001, 2008), respectively Section 2.2 of Kohl (2005) is returned.
-
- The OBRE, i.e., the optimal bias-robust (asymptotically linear) estimator;
- (terminology due to Hampel et al (1985)), expects an efficiency loss (at the
- ideal model) to be specified and then, according to an (asymptotic) Anscombe
- criterion computes the the bias bound achieving this efficiency loss.
-
- The MBRE, i.e., the most bias-robust (asymptotically linear) estimator;
- (terminology due to Hampel et al (1985)), uses the influence curve with
- minimal possible bias bound, hence minimaxes bias on these neighborhoods
- (in an infinitesimal sense)..
-
- Finite-sample and higher order results suggest that the asymptotically
- optimal procedure is to liberal. Using \code{fsCor} the radius can be
- modified - as a rule enlarged - to obtain a more conservative estimate.
- In case of normal location and scale there is function
- \code{\link[RobLox]{finiteSampleCorrection}} which returns a finite-sample
- corrected (enlarged) radius based on the results of large Monte-Carlo
- studies.
-
- The default value of argument \code{useLast} is set by the
- global option \code{kStepUseLast} which by default is set to
- \code{FALSE}. In case of general models \code{useLast}
- remains unchanged during the computations. However, if
- slot \code{CallL2Fam} of \code{IC} generates an object of
- class \code{"L2GroupParamFamily"} the value of \code{useLast}
- is changed to \code{TRUE}.
- Explicitly setting \code{useLast} to \code{TRUE} should
- be done with care as in this situation the influence curve
- is re-computed using the value of the one-step estimate
- which may take quite a long time depending on the model.
-
- If \code{useLast} is set to \code{TRUE} the computation of \code{asvar},
- \code{asbias} and \code{IC} is based on the k-step estimate.
-
- All these estimators are realized as wrappers to function \code{roptest}.
-
- Timings for the steps run through in these estimators are available
- in attributes \code{timings}, and for the step of the
- \code{kStepEstimator} in \code{kStepTimings}.
-
- One may also use the arguments \code{startCtrl}, \code{startICCtrl}, and
- \code{kStepCtrl} of function \code{\link{robest}}. This allows for individual
- settings of \code{E.argList}, \code{withEvalAsVar}, and
- \code{withMakeIC} for the different steps. If any of the three arguments
- \code{startCtrl}, \code{startICCtrl}, and \code{kStepCtrl} is used, the
- respective attributes set in the correspondig argument are used and, if
- colliding with arguments directly passed to the estimator function, the directly
- passed ones are ignored.
-
- Diagnostics on the involved integrations are available if argument
- \code{diagnostic} is \code{TRUE}. Then there are attributes \code{diagnostic}
- and \code{kStepDiagnostic} attached to the return value, which may be inspected
- and assessed through \code{\link[distrEx:distrExIntegrate]{showDiagnostic}} and
- \code{\link[distrEx:distrExIntegrate]{getDiagnostic}}.
-
-}
-\value{Object of class \code{"kStepEstimate"}. In addition, it has
- an attribute \code{"timings"} where computation time is stored.}
-\references{
- Kohl, M. (2005) \emph{Numerical Contributions to the Asymptotic Theory of Robustness}.
- Bayreuth: Dissertation.
-
- Kohl, M. and Ruckdeschel, P. (2010): R package distrMod:
- Object-Oriented Implementation of Probability Models.
- J. Statist. Softw. \bold{35}(10), 1--27
-
- Kohl, M. and Ruckdeschel, P., and Rieder, H. (2010):
- Infinitesimally Robust Estimation in General Smoothly Parametrized Models.
- \emph{Stat. Methods Appl.}, \bold{19}, 333--354.
-
- Rieder, H. (1994) \emph{Robust Asymptotic Statistics}. New York: Springer.
-
- Rieder, H., Kohl, M. and Ruckdeschel, P. (2008) The Costs of not Knowing
- the Radius. Statistical Methods and Applications \bold{17}(1) 13-40.
-
- Rieder, H., Kohl, M. and Ruckdeschel, P. (2001) The Costs of not Knowing
- the Radius. Appeared as discussion paper Nr. 81.
- SFB 373 (Quantification and Simulation of Economic Processes),
- Humboldt University, Berlin; also available under
- \url{www.uni-bayreuth.de/departments/math/org/mathe7/RIEDER/pubs/RR.pdf}
-}
-\author{Matthias Kohl \email{Matthias.Kohl at stamats.de},\cr
- Peter Ruckdeschel \email{peter.ruckdeschel at uni-oldenburg.de}}
-%\note{}
-\examples{
-#############################
-## 1. Binomial data
-#############################
-## generate a sample of contaminated data
-set.seed(123)
-ind <- rbinom(100, size=1, prob=0.05)
-x <- rbinom(100, size=25, prob=(1-ind)*0.25 + ind*0.9)
-
-## ML-estimate
-MLE.bin <- MLEstimator(x, BinomFamily(size = 25))
-## compute optimally robust estimators
-OMSE.bin <- OMSEstimator(x, BinomFamily(size = 25), steps = 3)
-MBRE.bin <- MBREstimator(x, BinomFamily(size = 25), steps = 3)
-estimate(MLE.bin)
-estimate(MBRE.bin)
-estimate(OMSE.bin)
-
-\donttest{ ## to reduce time load at CRAN tests
-RMXE.bin <- RMXEstimator(x, BinomFamily(size = 25), steps = 3)
-OBRE.bin <- OBREstimator(x, BinomFamily(size = 25), steps = 3)
-estimate(RMXE.bin)
-estimate(OBRE.bin)
-}
-\donttest{ ## to reduce time load at CRAN tests
-#############################
-## 2. Poisson data
-#############################
-
-## Example: Rutherford-Geiger (1910); cf. Feller~(1968), Section VI.7 (a)
-x <- c(rep(0, 57), rep(1, 203), rep(2, 383), rep(3, 525), rep(4, 532),
- rep(5, 408), rep(6, 273), rep(7, 139), rep(8, 45), rep(9, 27),
- rep(10, 10), rep(11, 4), rep(12, 0), rep(13, 1), rep(14, 1))
-
-## ML-estimate
-MLE.pois <- MLEstimator(x, PoisFamily())
-OBRE.pois <- OBREstimator(x, PoisFamily(), steps = 3)
-OMSE.pois <- OMSEstimator(x, PoisFamily(), steps = 3)
-MBRE.pois <- MBREstimator(x, PoisFamily(), steps = 3)
-RMXE.pois <- RMXEstimator(x, PoisFamily(), steps = 3)
-estimate(MLE.pois)
-estimate(OBRE.pois)
-estimate(RMXE.pois)
-estimate(MBRE.pois)
-estimate(OMSE.pois)
-}
-
-\donttest{ ## to reduce time load at CRAN tests
-#############################
-## 3. Normal (Gaussian) location and scale
-#############################
-## 24 determinations of copper in wholemeal flour
-library(MASS)
-data(chem)
-
-MLE.n <- MLEstimator(chem, NormLocationScaleFamily())
-MBRE.n <- MBREstimator(chem, NormLocationScaleFamily(), steps = 3)
-OMSE.n <- OMSEstimator(chem, NormLocationScaleFamily(), steps = 3)
-OBRE.n <- OBREstimator(chem, NormLocationScaleFamily(), steps = 3)
-RMXE.n <- RMXEstimator(chem, NormLocationScaleFamily(), steps = 3)
-
-estimate(MLE.n)
-estimate(MBRE.n)
-estimate(OMSE.n)
-estimate(OBRE.n)
-estimate(RMXE.n)
-}
-}
-
-\seealso{ \code{\link{roptest}}, \code{\link{robest}},
- \code{\link[RobLox]{roblox}},
- \code{\link[distrMod]{L2ParamFamily-class}}
- \code{\link[RobAStBase]{UncondNeighborhood-class}},
- \code{\link[distrMod]{RiskType-class}} }
-\concept{k-step construction}
-\concept{optimally robust estimation}
-\keyword{robust}
+\name{RMXEOMSEMBREOBRE}
+\alias{RMXEstimator}
+\alias{OMSEstimator}
+\alias{OBREstimator}
+\alias{MBREstimator}
+\title{ Optimally robust estimation: RMXE, OMSE, MBRE, and OBRE }
+\description{
+ These are wrapper functions to 'roptest' to compute
+ optimally robust estimates, more specifically RMXEs, OMSEs, MBREs, and OBREs,
+ for L2-differentiable parametric families via k-step construction.
+}
+\usage{
+RMXEstimator(x, L2Fam, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
+ steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
+ OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
+ withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
+ IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
+ withICList = getRobAStBaseOption("withICList"),
+ withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
+ initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
+ withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
+ withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
+ diagnostic = FALSE)
+OMSEstimator(x, L2Fam, eps=0.5, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
+ steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
+ OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
+ withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
+ IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
+ withICList = getRobAStBaseOption("withICList"),
+ withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
+ initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
+ withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
+ withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
+ diagnostic = FALSE)
+OBREstimator(x, L2Fam, eff=0.95, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
+ steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
+ OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
+ withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
+ IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
+ withICList = getRobAStBaseOption("withICList"),
+ withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
+ initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
+ withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
+ withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
+ diagnostic = FALSE)
+MBREstimator(x, L2Fam, fsCor = 1, initial.est, neighbor = ContNeighborhood(),
+ steps = 1L, distance = CvMDist, startPar = NULL, verbose = NULL,
+ OptOrIter = "iterate", useLast = getRobAStBaseOption("kStepUseLast"),
+ withUpdateInKer = getRobAStBaseOption("withUpdateInKer"),
+ IC.UpdateInKer = getRobAStBaseOption("IC.UpdateInKer"),
+ withICList = getRobAStBaseOption("withICList"),
+ withPICList = getRobAStBaseOption("withPICList"), na.rm = TRUE,
+ initial.est.ArgList, ..., withLogScale = TRUE, ..withCheck=FALSE,
+ withTimings = FALSE, withMDE = NULL, withEvalAsVar = NULL,
+ withMakeIC = FALSE, modifyICwarn = NULL, E.argList = NULL,
+ diagnostic = FALSE)
+}
+\arguments{
+ \item{x}{ sample }
+ \item{L2Fam}{ object of class \code{"L2ParamFamily"} }
+ \item{eff}{ positive real (0 <= \code{eff} <= 1): amount of asymptotic
+ efficiency loss in the ideal model. See details below. }
+ \item{eps}{ positive real (0 < \code{eps} <= 0.5): amount of gross errors.
+ See details below. }
+ \item{fsCor}{ positive real: factor used to correct the neighborhood radius;
+ see details. }
+ \item{initial.est}{ initial estimate for unknown parameter. If missing
+ minimum distance estimator is computed. }
+ \item{neighbor}{ object of class \code{"UncondNeighborhood"} }
+ \item{steps}{ positive integer: number of steps used for k-steps construction }
+ \item{distance}{ distance function used in \code{MDEstimator}, which in turn
+ is used as (default) starting estimator. }
+ \item{startPar}{ initial information used by \code{optimize} resp. \code{optim};
+ i.e; if (total) parameter is of length 1, \code{startPar} is
+ a search interval, else it is an initial parameter value; if \code{NULL}
+ slot \code{startPar} of \code{ParamFamily} is used to produce it;
+ in the multivariate case, \code{startPar} may also be of class \code{Estimate},
+ in which case slot \code{untransformed.estimate} is used.}
+ \item{verbose}{ logical: if \code{TRUE}, some messages are printed }
+ \item{useLast}{ which parameter estimate (initial estimate or
+ k-step estimate) shall be used to fill the slots \code{pIC},
+ \code{asvar} and \code{asbias} of the return value. }
+ \item{OptOrIter}{character; which method to be used for determining Lagrange
+ multipliers \code{A} and \code{a}: if (partially) matched to \code{"optimize"},
+ \code{getLagrangeMultByOptim} is used; otherwise: by default, or if matched to
+ \code{"iterate"} or to \code{"doubleiterate"},
+ \code{getLagrangeMultByIter} is used. More specifically,
+ when using \code{getLagrangeMultByIter}, and if argument \code{risk} is of
+ class \code{"asGRisk"}, by default and if matched to \code{"iterate"}
+ we use only one (inner) iteration, if matched to \code{"doubleiterate"}
+ we use up to \code{Maxiter} (inner) iterations.}
+ \item{withUpdateInKer}{if there is a non-trivial trafo in the model with matrix \eqn{D}, shall
+ the parameter be updated on \eqn{{\rm ker}(D)}{ker(D)}?}
+ \item{IC.UpdateInKer}{if there is a non-trivial trafo in the model with matrix \eqn{D},
+ the IC to be used for this; if \code{NULL} the result of \code{getboundedIC(L2Fam,D)} is taken;
+ this IC will then be projected onto \eqn{{\rm ker}(D)}{ker(D)}.}
+ \item{withPICList}{logical: shall slot \code{pICList} of return value
+ be filled?}
+ \item{withICList}{logical: shall slot \code{ICList} of return value
+ be filled?}
+ \item{na.rm}{logical: if \code{TRUE}, the estimator is evaluated at \code{complete.cases(x)}.}
+ \item{initial.est.ArgList}{a list of arguments to be given to argument \code{start} if the latter
+ is a function; this list by default already starts with two unnamed items,
+ the sample \code{x}, and the model \code{L2Fam}.}
+ \item{\dots}{ further arguments }
+ \item{withLogScale}{logical; shall a scale component (if existing and found
+ with name \code{scalename}) be computed on log-scale and backtransformed
+ afterwards? This avoids crossing 0.}
+ \item{..withCheck}{logical: if \code{TRUE}, debugging info is issued.}
+ \item{withTimings}{logical: if \code{TRUE}, separate (and aggregate)
+ timings for the three steps evaluating the starting value, finding
+ the starting influence curve, and evaluating the k-step estimator is
+ issued.}
+ \item{withMDE}{ logical or \code{NULL}: Shall a minimum distance estimator be used as
+ starting estimator---in addition to the function given in slot
+ \code{startPar} of the L2 family? If \code{NULL} (default), the content
+ of slot \code{.withMDE} in the L2 family is used instead to take
+ this decision.}
+ \item{withEvalAsVar}{logical or \code{NULL}: if \code{TRUE} (default), tells R
+ to evaluate the asymptotic variance or if \code{FALSE} just to produces a call
+ to do so. If \code{withEvalAsVar} is \code{NULL} (default), the content
+ of slot \code{.withEvalAsVar} in the L2 family is used instead to take
+ this decision.}
+ \item{withMakeIC}{logical; if \code{TRUE} the [p]IC is passed through
+ \code{makeIC} before return.}
+ \item{modifyICwarn}{logical: should a (warning) information be added if
+ \code{modifyIC} is applied and hence some optimality information could
+ no longer be valid? Defaults to \code{NULL} in which case this value
+ is taken from \code{RobAStBaseOptions}.}
+ \item{E.argList}{\code{NULL} (default) or a list of arguments to be passed
+ to calls to \code{E} from (a) \code{MDEstimator}
+ (here this additional argument is only used if
+ \code{initial.est} is missing), (b) \code{getStartIC},
+ and (c) \code{kStepEstimator}. Potential clashes with
+ arguments of the same name in \code{\dots} are resolved by inserting
+ the items of argument list \code{E.argList} as named items, so
+ in case of collisions the item of \code{E.argList} overwrites the
+ existing one from \code{\dots}.}
+ \item{diagnostic}{ logical; if \code{TRUE},
+ diagnostic information on the performed integrations is gathered and
+ shipped out as an attribute \code{diagnostic} of the return value
+ of the estimators. }
+}
+\details{
+ The functions compute optimally robust estimator for a given L2 differentiable
+ parametric family; more specifically they are RMXEs, OMSEs, MBREs, and OBREs.
+ The computation uses a k-step construction with an
+ appropriate initial estimate; cf. also \code{\link[RobAStBase]{kStepEstimator}}.
+ Valid candidates are e.g. Kolmogorov(-Smirnov) or von Mises minimum
+ distance estimators (default); cf. Rieder (1994) and Kohl (2005).
+
+ For OMSE, i.e., the asymptotically linear estimator with minimax mean squared
+ error on this neighborhood of given size, the amount of gross errors
+ (contamination) is assumed to be known, and is specified by \code{eps}.
+ The radius of the corresponding infinitesimal
+ contamination neighborhood is obtained by multiplying \code{eps}
+ by the square root of the sample size.
+
+ If the amount of gross errors (contamination) is unknown, RMXE should be used,
+ i.e., the radius-minimax estimator in the sense of
+ Rieder et al. (2001, 2008), respectively Section 2.2 of Kohl (2005) is returned.
+
+ The OBRE, i.e., the optimal bias-robust (asymptotically linear) estimator;
+ (terminology due to Hampel et al (1985)), expects an efficiency loss (at the
+ ideal model) to be specified and then, according to an (asymptotic) Anscombe
+ criterion computes the the bias bound achieving this efficiency loss.
+
+ The MBRE, i.e., the most bias-robust (asymptotically linear) estimator;
+ (terminology due to Hampel et al (1985)), uses the influence curve with
+ minimal possible bias bound, hence minimaxes bias on these neighborhoods
+ (in an infinitesimal sense)..
+
+ Finite-sample and higher order results suggest that the asymptotically
+ optimal procedure is to liberal. Using \code{fsCor} the radius can be
+ modified - as a rule enlarged - to obtain a more conservative estimate.
+ In case of normal location and scale there is function
+ \code{\link[RobLox]{finiteSampleCorrection}} which returns a finite-sample
+ corrected (enlarged) radius based on the results of large Monte-Carlo
+ studies.
+
+ The default value of argument \code{useLast} is set by the
+ global option \code{kStepUseLast} which by default is set to
+ \code{FALSE}. In case of general models \code{useLast}
+ remains unchanged during the computations. However, if
+ slot \code{CallL2Fam} of \code{IC} generates an object of
+ class \code{"L2GroupParamFamily"} the value of \code{useLast}
+ is changed to \code{TRUE}.
+ Explicitly setting \code{useLast} to \code{TRUE} should
+ be done with care as in this situation the influence curve
+ is re-computed using the value of the one-step estimate
+ which may take quite a long time depending on the model.
+
+ If \code{useLast} is set to \code{TRUE} the computation of \code{asvar},
+ \code{asbias} and \code{IC} is based on the k-step estimate.
+
+ All these estimators are realized as wrappers to function \code{roptest}.
+
+ Timings for the steps run through in these estimators are available
+ in attributes \code{timings}, and for the step of the
+ \code{kStepEstimator} in \code{kStepTimings}.
+
+ One may also use the arguments \code{startCtrl}, \code{startICCtrl}, and
+ \code{kStepCtrl} of function \code{\link{robest}}. This allows for individual
+ settings of \code{E.argList}, \code{withEvalAsVar}, and
+ \code{withMakeIC} for the different steps. If any of the three arguments
+ \code{startCtrl}, \code{startICCtrl}, and \code{kStepCtrl} is used, the
+ respective attributes set in the correspondig argument are used and, if
+ colliding with arguments directly passed to the estimator function, the directly
+ passed ones are ignored.
+
+ Diagnostics on the involved integrations are available if argument
+ \code{diagnostic} is \code{TRUE}. Then there are attributes \code{diagnostic}
+ and \code{kStepDiagnostic} attached to the return value, which may be inspected
+ and assessed through \code{\link[distrEx:distrExIntegrate]{showDiagnostic}} and
+ \code{\link[distrEx:distrExIntegrate]{getDiagnostic}}.
+
+}
+\value{Object of class \code{"kStepEstimate"}. In addition, it has
+ an attribute \code{"timings"} where computation time is stored.}
+\references{
+ Kohl, M. (2005) \emph{Numerical Contributions to the Asymptotic Theory of Robustness}.
+ Bayreuth: Dissertation.
+
+ Kohl, M. and Ruckdeschel, P. (2010): R package distrMod:
+ Object-Oriented Implementation of Probability Models.
+ J. Statist. Softw. \bold{35}(10), 1--27
+
+ Kohl, M. and Ruckdeschel, P., and Rieder, H. (2010):
+ Infinitesimally Robust Estimation in General Smoothly Parametrized Models.
+ \emph{Stat. Methods Appl.}, \bold{19}, 333--354.
+
+ Rieder, H. (1994) \emph{Robust Asymptotic Statistics}. New York: Springer.
+
+ Rieder, H., Kohl, M. and Ruckdeschel, P. (2008) The Costs of not Knowing
+ the Radius. Statistical Methods and Applications \bold{17}(1) 13-40.
+
+ Rieder, H., Kohl, M. and Ruckdeschel, P. (2001) The Costs of not Knowing
+ the Radius. Appeared as discussion paper Nr. 81.
+ SFB 373 (Quantification and Simulation of Economic Processes),
+ Humboldt University, Berlin; also available under
+ \url{https://dx.doi.org/10.18452/3638}
+}
+\author{Matthias Kohl \email{Matthias.Kohl at stamats.de},\cr
+ Peter Ruckdeschel \email{peter.ruckdeschel at uni-oldenburg.de}}
+%\note{}
+\examples{
+#############################
+## 1. Binomial data
+#############################
+## generate a sample of contaminated data
+set.seed(123)
+ind <- rbinom(100, size=1, prob=0.05)
+x <- rbinom(100, size=25, prob=(1-ind)*0.25 + ind*0.9)
+
+## ML-estimate
+MLE.bin <- MLEstimator(x, BinomFamily(size = 25))
+## compute optimally robust estimators
+OMSE.bin <- OMSEstimator(x, BinomFamily(size = 25), steps = 3)
+MBRE.bin <- MBREstimator(x, BinomFamily(size = 25), steps = 3)
+estimate(MLE.bin)
+estimate(MBRE.bin)
+estimate(OMSE.bin)
+
+\donttest{ ## to reduce time load at CRAN tests
+RMXE.bin <- RMXEstimator(x, BinomFamily(size = 25), steps = 3)
+OBRE.bin <- OBREstimator(x, BinomFamily(size = 25), steps = 3)
+estimate(RMXE.bin)
+estimate(OBRE.bin)
+}
+\donttest{ ## to reduce time load at CRAN tests
+#############################
+## 2. Poisson data
+#############################
+
+## Example: Rutherford-Geiger (1910); cf. Feller~(1968), Section VI.7 (a)
+x <- c(rep(0, 57), rep(1, 203), rep(2, 383), rep(3, 525), rep(4, 532),
+ rep(5, 408), rep(6, 273), rep(7, 139), rep(8, 45), rep(9, 27),
+ rep(10, 10), rep(11, 4), rep(12, 0), rep(13, 1), rep(14, 1))
+
+## ML-estimate
+MLE.pois <- MLEstimator(x, PoisFamily())
+OBRE.pois <- OBREstimator(x, PoisFamily(), steps = 3)
+OMSE.pois <- OMSEstimator(x, PoisFamily(), steps = 3)
+MBRE.pois <- MBREstimator(x, PoisFamily(), steps = 3)
[TRUNCATED]
To get the complete diff run:
svnlook diff /svnroot/robast -r 1243
More information about the Robast-commits
mailing list