Breakdown points for redescending m-estimates of location

Breakdown point is one measure of the robustness of an estimate. This paper discusses some unusual properties of the breakdown points of M-estimates of location.     

The calculation of outlier detection statistics

This paper presents a routine that calculates four outlier detection statistics. The routine determines a series of points that are identified as possible outliers, and calculates the values that can be used to test them. These values can be used in an iterative procedure to detect multiple outliers.     

A note on james type robust estimators

M-estimation of a single parameter of the life time distribution is considered based on independent and identically distributed survival data which may be randomly censored. The most robust and the optimal robust M-estimators of the location parameters of the survival time distribution are derived within a class considered in James (1986) as well as for the general unrestricted class. The properties of the estimators corresponding to the above two classes are discussed. A data set is used to illustrate the usefulness of the optimal robust estimators for the parameter of extreme value distribution.     

Nonparametric geometric outlier detection

Outlier detection is a major topic in robust statistics due to the high practical significance of anomalous observations. Many existing methods, however, either are parametric or cease to perform well when the data are far from linearly structured. In this paper, we propose a quantity, Delaunay outlyingness, that is a nonparametric outlyingness score applicable to data with complicated structure. The approach is based on a well‐known triangulation of the sample, which seems to reflect the sparsity of the pointset to different directions in a useful way. We derive results on the asymptotic behavior of Delaunay outlyingness in case of a sufficiently simple set of observations. Simulations and an application to empirical data are also discussed.     

The heuristic approach in finding initial values for minimum density power divergence estimators

It is well known that in the presence of outliers the maximum likelihood estimates are very unstable. In these situations, an alternative is resorting to the estimators based on the minimum density power divergence criterion for which feasible, computationally closed-form expressions can be derived, so that solutions can be achieved by any standard nonlinear optimization code. But since the function to be minimized is often ill-behaved, the convergence of the algorithm to optimal solutions strongly depends on the choice of the configuration of the initial values. A new procedure based on a heuristic local search approach is introduced in order to survey the parameters space and hence obtaining an accurate set of starting guesses for the gradient-method minimization routine.     

Features and performance of some outlier detection methods

A review of several statistical methods that are currently in use for outlier identification is presented, and their performances are compared theoretically for typical statistical distributions of experimental data, considering values derived from the distribution of extreme order statistics as reference terms. A simple modification of a popular, broadly used method based upon box-plot is introduced, in order to overcome a major limitation concerning sample size. Examples are presented concerning exploitation of methods considered on two data sets: a historical one concerning evaluation of an astronomical constant performed by a number of leading observatories and a substantial database pertaining to an ongoing investigation on absolute measurement of gravity acceleration, exhibiting peculiar aspects concerning outliers. Some problems related to outlier treatment are examined, and the requirement of both statistical analysis and expert opinion for proper outlier management is underlined.     

Bounded influence estimation for regression and scale

Bednarski and Müller [Optimal bounded influence regression and scale M-estimators in the context of experimental design, Statistics 35 (2001), pp. 349–369] introduced a class of bounded influence M estimates for the simultaneous estimation of regression and scale in the linear model with normal errors by solving the corresponding normal location and scale problem at each design point. This limits the proposal to regressor distributions with finite support. Based on their approach, we propose a slightly extended class of M estimates that is not restricted to finite support and is numerically easier to handle. Moreover, we employ the even more general class of asymptotically linear (AL) estimators which, in addition, is not restricted to normal errors. The superiority of AL estimates is demonstrated by numerical comparisons of the maximum asymptotic mean-squared error over infinitesimal contamination neighbourhoods.     

Estimation and testing under exchangeable exponential model with a single outlier

Several estimators of mean of an exponential distribution, when an unidentified single outlier in a sample of size n is present* are discussed. It is assumed that n?1 of these observations have a mean σ, While one could have a mean σ/α. The estimation of σ has been considered in detail with some reference to the estimation of α. Finally, tests of hypotheses about σare briefly mentioned.     

Some properties of generalized ridge estimators

Exact expressions, in the form of infinite series expansions, are given for the first and second moments of two well known generalized ridge estimators. These series expansions are then evaluated using recursive formulas and computations are verified using approximations. Results are presented for the relative mean square error and bias of these estimators as well as their relative efficiency with respect to least squares.     

Optimal,nearly optimal and robust procedures for the exponential distribution with relative linear exponential loss

In this paper, the problem of sequentially estimating the mean of the exponential distribution with relative linear exponential loss and fixed cost for each observation is considered within the Bayesian framework. An optimal procedure with a deterministic stopping rule is derived. Since the corresponding value of the optimal deterministic stopping rule cannot be obtained directly, an approximate optimal deterministic stopping rule and an asymptotically pointwise optimal rule are proposed. In addition, we propose a robust procedure with a deterministic stopping rule, which does not depend on the parameters of the prior distribution. All of the proposed procedures are shown to be asymptotically optimal. Some numerical studies are conducted to investigate the performances of the proposed procedures. A real data set is provided to illustrate the use of the proposed procedures.     

Unbiasedness of two-stage estimation and prediction procedures for mixed linear models

The traditional method for estimating or predicting linear combinations of the fixed effects and realized values of the random effects in mixed linear models is first to estimate the variance components and then to proceed as if the estimated values of the variance components were the true values. This two-stage procedure gives unbiased estimators or predictors of the linear combinations provided the data vector is symmetrically distributed about its expected value and provided the variance component estimators are translation-invariant and are even functions of the data vector. The standard procedures for estimating the variance components yield even, translation-invariant estimators.     

Goodness-of-fit-based outlier detection for Phase I monitoring

ABSTRACT

Nonparametric charts are useful in statistical process control when there is a lack of or limited knowledge about the underlying process distribution. Most existing approaches in the literature of Phase I monitoring assume that outliers have the same distributions as the in-control sample but only differ in location or scale parameters, they may not be effective with distributional changes. This article develops a new procedure based on the integration of the classical Anderson–Darling goodness-of-fit test and the stepwise isolation method. Our proposed procedure is efficient in detecting potential shifts in location, scale, or shape, and thus it offers robust protection against variation in various underlying distributions. The finite sample performance of our method is evaluated through simulations and is compared with that of available outlier detection methods for Phase I monitoring.     

On the performance of some subset selection procedures

In this paper, we restrict attention to the problem of subset selection of normal populations. The approaches and results of some previous comparison studies of subset selection procedures are discussed briefly. And then the result of a new Monte Carlo study comparing the performance of two classical procedures and the Bayes procedure is presented.     

The internal norm approach to influence diagnostics

Cook and Weisberg (1982) describe the external and internal norm approaches to assessing the influence of a subset of data on least squares regression estimates. External norms base influence measurement on the repeated sampling theory of the assumed model, while internal norm measures judge the influence of a size-k subset relative to all size-k subsets within the given data. Although intuitively appealing, intemal norms have been largely ignored in favor of external norms due to computational considerations. The purpose of this article is to present the internal norm approach as a viable alternative to external norm influence measurement. In addition to discussing conceptual and computational issues, empirical evidence is provided to show that the internal norm interpretation of influence is different from that of its external counterparts. Finally, comparisons are drawn between external calibration and internal scaling for evaluating influence measure values.     

A domain outlier robust design and smooth estimation approach

Outliers that commonly occur in business sample surveys can have large impacts on domain estimates. The authors consider an outlier‐robust design and smooth estimation approach, which can be related to the so‐called “Surprise stratum” technique [Kish, “Survey Sampling,” Wiley, New York (1965)]. The sampling design utilizes a threshold sample consisting of previously observed outliers that are selected with probability one, together with stratified simple random sampling from the rest of the population. The domain predictor is an extension of the Winsorization‐based estimator proposed by Rivest and Hidiroglou [Rivest and Hidiroglou, “Outlier Treatment for Disaggregated Estimates,” in “Proceedings of the Section on Survey Research Methods,” American Statistical Association (2004), pp. 4248–4256], and is similar to the estimator for skewed populations suggested by Fuller [Fuller, Statistica Sinica 1991;1:137–158]. It makes use of a domain Winsorized sample mean plus a domain‐specific adjustment of the estimated overall mean of the excess values on top of that. The methods are studied in theory from a design‐based perspective and by simulations based on the Norwegian Research and Development Survey data. Guidelines for choosing the threshold values are provided. The Canadian Journal of Statistics 39: 147–164; 2011 © 2010 Statistical Society of Canada     

The successive raising estimator and its relation with the ridge estimator

The raised estimators are used to reduce collinearity in linear regression models by raising a column in the experimental data matrix which may be nearly linear with the other columns. The raising procedure has two components, namely stretching and rotating, which we can analyze separately. We give the relationship between the raised estimators and the classical ridge estimators. Using a case study, we show how to determine the perturbation parameter for the raised estimators by controlling the amount of precision to be retained in the original data.     

Percentage points and power calculations for outlier tests in a regression situation

The usefulness of an extra sum of squares statistics Q_K for detecting K outliers has been discussed previously in the context of two-way tables. (See Gentleman and Wilk, 1975a, 1975b; John and Draper 1978; and Draper and John, 1980,) That work is extended here to straight line regression situations arising from, and motivated by, a specific set of research data. Percentage points for the appropriate test statistics are obtained by simulation, and approximations for these percentage points are suggested. Power calculations made for various designs and outlier situations are briefly summarized.     

Consistent procedures for multiclass classification of discrete diffusion paths

The recent advent of modern technology has generated a large number of datasets which can be frequently modeled as functional data. This paper focuses on the problem of multiclass classification for stochastic diffusion paths. In this context we establish a closed formula for the optimal Bayes rule. We provide new statistical procedures which are built either on the plug-in principle or on the empirical risk minimization principle. We show the consistency of these procedures under mild conditions. We apply our methodologies to the parametric case and illustrate their accuracy with a simulation study through examples.     

Estimation and selection procedures in regression: An L1 approach

The authors consider the problem of estimating a regression function go involving several variables by the closest functional element of a prescribed class _G that is closest to it in the L₁ norm. They propose a new estimator ? based on independent observations and give explicit finite sample bounds for the L₁distance between ?g and go. They apply their estimation procedure to the problem of selecting the smoothing parameter in nonparametric regression.     

Stable estimation of location parameter -nonasymptotic approach

Let X₁:, X₂:, …, X_n be iidrv's with cdf F?, F?(x)=F (x-θ), R. Let T be an equivariant median-unbiased estimator of θ. Let πε(F)={G = (1 -ε) F+εH, H any cdf} and let M(G, T) be a median of T if X₁ has cdf G. The oscillation of the bias of T, defined as

Bε(T)=sup (M(G₁ T) :G₁,G₂:∈π_σ:(F)} ,is considered and the estimator with the smallest B$epsi;(T) is explicitly constructed