Large adaptive estimation in linear regression model. I. Consistency
On optimality of the orthogonal block design
In the paper a usual block design with treatment effects fixed and block effects random is considered. To compare experimental design the asymptotic covariance matrix of a robust estimator proposed by Bednarski and Zontek (1996) for simultaneous estimation of shift and scale parameters is used. Asymptotically A- and D- optimal block designs in the class of designs with bounded block sizes are characterized.
Permanents, order statistics, outliers, and robustness.
In this paper, we consider order statistics and outlier models, and focus primarily on multiple-outlier models and associated robustness issues. We first synthesise recent developments on order statistics arising from independent and non-identically distributed random variables based primarily on the theory of permanents. We then highlight various applications of these results in evaluating the robustness properties of several linear estimators when multiple outliers are possibly present in the...
Problèmes de méthodologie statistique. I. Introduction à l'étude de la robustesse des méthodes usuelles d'inférence statistique
La connaissance de la robustesse des méthodes usuelles d'inférence statistique vis-à-vis de divers types d'écarts par rapport au modèle de base est essentielle pour leur bonne utilisation. Dans cet article sont exposés un certain nombre de résultats (pour la plupart classiques. mais parfois mal connus) concernant la robustesse de ces méthodes vis-à-vis de la non-normalité (pour les comparaisons de moyennes, puis pour les comparaisons de variances), vis-à-vis de la non-équidistribution et de la non-indépendance,...
Problèmes de méthodologie statistique. II. —Étude d'un conflit robustesse-efficacité dans le problème de la comparaison de deux moyennes (groupes indépendants)
Le conflit robustesse-efficacité se trouve posé notamment chaque fois que l'on a à choisir entre deux méthodes d'inférence statistique dont l'une est privilégiée sous un modèle plus spécifique et l'autre sous un modèle plus général. Ce conflit est étudié dans le cas de la comparaison de deux moyennes (groupes indépendants), à propos du choix entre le modèle (spécifique) postulant l'égalité des variances intra-groupes et le modèle (général) à variances quelconques. On montre que le choix n'est crucial...
Rank tests of symmetry and R-estimation of location parameter under measurement errors
This paper deals with the hypotheses of symmetry of distributions with respect to a location parameter when the response variables are subject to measurement errors. Rank tests of hypotheses about the location parameter and the related R-estimators are studied in an asymptotic set up. It is shown, when and under what conditions, these rank tests and R-estimators can be used effectively, and the effect of measurement errors on the power of the test and on the efficiency of the R-estimators is indicated....
Redescending M-estimators in regression analysis, cluster analysis and image analysis
We give a review on the properties and applications of M-estimators with redescending score function. For regression analysis, some of these redescending M-estimators can attain the maximum breakdown point which is possible in this setup. Moreover, some of them are the solutions of the problem of maximizing the efficiency under bounded influence function when the regression coefficient and the scale parameter are estimated simultaneously. Hence redescending M-estimators satisfy several outlier robustness...
Réponse de M. Cramer au commentaire de M. Rey
Robust estimation based on spacings in weighted exponential models
Using Zieliński's (1977, 1983) formalization of robustness Błażej (2007) obtained uniformly most bias-robust estimates (UMBREs) of the scale parameter for some statistical models (including the exponential model), in a class of linear functions of order statistics, when violations of the models are generated by weight functions. In this paper the UMBRE of the scale parameter, based on spacings, in two weighted exponential models is derived. Extensions of results of Bartoszewicz (1986, 1987) are...
Robustness regions for measures of risk aggregation
One of risk measures’ key purposes is to consistently rank and distinguish between different risk profiles. From a practical perspective, a risk measure should also be robust, that is, insensitive to small perturbations in input assumptions. It is known in the literature [14, 39], that strong assumptions on the risk measure’s ability to distinguish between risks may lead to a lack of robustness. We address the trade-off between robustness and consistent risk ranking by specifying the regions in...
Robustness to non-normality of various tests for the one-sample location problem.
Scatter halfspace depth: Geometric insights
Scatter halfspace depth is a statistical tool that allows one to quantify the fitness of a candidate covariance matrix with respect to the scatter structure of a probability distribution. The depth enables simultaneous robust estimation of location and scatter, and nonparametric inference on these. A handful of remarks on the definition and the properties of the scatter halfspace depth are provided. It is argued that the currently used notion of this depth is well suited especially for symmetric...
Some Diagnostic Tools in Robust Econometrics
Highly robust statistical and econometric methods have been developed not only as a diagnostic tool for standard methods, but they can be also used as self-standing methods for valid inference. Therefore the robust methods need to be equipped by their own diagnostic tools. This paper describes diagnostics for robust estimation of parameters in two econometric models derived from the linear regression. Both methods are special cases of the generalized method of moments estimator based on implicit...
The infinitesimal robustness of tests against dependence
The robustness against dependence of nonparametric tests for the two-sample location problem
Nonparametric tests for the two-sample location problem are investigated. It is shown that the supremum of the size of any test can be arbitrarily close to 1. None of these tests is most robust against dependence.
Toward the best constant factor for the Rademacher-Gaussian tail comparison
It is proved that the best constant factor in the Rademacher-Gaussian tail comparison is between two explicitly defined absolute constants c1 and c2 such that c2≈1.01 c1. A discussion of relative merits of this result versus limit theorems is given.