Breakdown points and variation exponents of robust \(M\)-estimators in linear models. (English) Zbl 0959.62029

Mizera, Ivan; Müller, Christine H.

Ann. Stat. 27, No. 4, 1164-1177 (1999).

Summary: The breakdown point behavior of \(M\)-estimators in linear models with fixed designs, arising from planned experiments or qualitative factors, is characterized. Particularly, this behavior at fixed designs is quite different from that at designs which can be corrupted by outliers, the situation prevailing in the literature. For fixed designs, the breakdown points of robust \(M\)-estimators (those with bounded derivative of the score function), depend on the design and the variation exponent (index) of the score function.
This general result implies that the highest breakdown point within all regression equivariant estimators can be attained also by certain \(M\)-estimators: those with slowly varying score function, like the Cauchy or slash maximum likelihood estimator. The \(M\)-estimators with variation exponent greater than 0, like the \(L_1\) or Huber estimator, exhibit a considerably worse breakdown point behavior.

MSC:

62F35	Robustness and adaptive procedures (parametric inference)
62F10	Point estimation
62J05	Linear regression; mixed models
62J10	Analysis of variance and covariance (ANOVA)
62K99	Design of statistical experiments

Keywords:

L1 estimator; linear model; M-estimator; regular variation; breakdown point

References:

[1]	Bingham, N. H., Goldie, C. M. and Teugels, J. L. (1987). Regular Variation. Cambridge Univ. Press. · Zbl 0617.26001
[2]	Copas, J. B. (1975). On the unimodality of the likelihood for the Cauchy distribution. Biometrika 62 701-704. JSTOR: · Zbl 0321.62037 · doi:10.1093/biomet/62.3.701
[3]	Ellis, S. P. and Morgenthaler, S. (1992). Leverage and breakdown in L1 regression. J. Amer. Statist. Assoc. 87 143-148. JSTOR: · Zbl 0781.62101 · doi:10.2307/2290462
[4]	Gabrielsen, G. (1982). On the unimodality of the likelihood for the Cauchy distribution: some comments. Biometrika 69 677-678. JSTOR: · Zbl 0499.62031 · doi:10.1093/biomet/69.3.677
[5]	Hampel, F. R., Ronchetti, E. M., Rousseeuw, P. J. and Stahel, W. A. (1986). Robust StatisticsThe Approach Based on Influence Functions. Wiley, New York. · Zbl 0593.62027
[6]	He, X., Jure cková, J., Koenker, R. and Portnoy, S. (1990). Tail behavior of regression estimators and their breakdown points. Econometrica 58 1195-1214. · Zbl 0745.62030 · doi:10.2307/2938306
[7]	Hoaglin, D. C., Mosteller, F. and Tukey, J. W. (eds.) (1983). Understanding Robust and Exploratory Data Analysis. Wiley, New York. · Zbl 0599.62007
[8]	Huber, P. J. (1981). Robust Statistics. Wiley, New York. · Zbl 0536.62025
[9]	Huber, P. J. (1984). Finite sample breakdown of Mand P-estimators. Ann. Statist. 12 119-126. · Zbl 0557.62034 · doi:10.1214/aos/1176346396
[10]	Jure cková, J. (1981). Tail-behavior of location estimators. Ann. Statist. 9 578-585. · Zbl 0476.62032 · doi:10.1214/aos/1176345461
[11]	Lange, K. L., Little R. J. A. and Taylor, J. M. G. (1989). Robust statistical modeling using the t distribution. J. Amer. Statist. Assoc. 84 881-896. JSTOR: · doi:10.2307/2290063
[12]	Maronna, R. A., Bustos, O. H. and Yohai, V. J. (1979). Biasand efficiency-robustness of general M-estimators for regression with random carriers. In Smoothing Techniques for Curve Estimation. Lecture Notes in Math. 757 91-116. Springer, Berlin. · Zbl 0416.62050 · doi:10.1007/BFb0098492
[13]	Mili, L. and Coakley, C. W. (1993). Robust estimation in structured linear regression. Ann. Statist. 24 2593-2607. · Zbl 0867.62040 · doi:10.1214/aos/1032181171
[14]	Mizera, I. (1994). On consistent M-estimators: tuning constants, unimodality and breakdown. Kybernetika 30 289-300. · Zbl 0815.62013
[15]	Mizera, I. (1996). Weak continuity of redescending M-estimators of location with an unbounded objective function. In PROBASTAT ’94, Proceedings of the Second International Conference on Mathematical Statistics (A. Pázman and V. Witkovský, eds.) 343-347. · Zbl 0919.62024
[16]	Mizera, I. and M üller, Ch. H. (1996). Breakdown points and variation exponents of robust Mestimators in linear models. Preprint A-22-96, Freie Univ. Berlin, Fachbereich Mathematik und Informatik, Ser. A (Mathematik). Available as No. 22 at ftp://ftp.math.fuberlin.de/pub/math/publ/pre/1996/index.html.
[17]	Morgenthaler, S. and Tukey, J. W. (1991). Configural Polysampling. Wiley, New York. · Zbl 0756.62016
[18]	M üller, Ch. H. (1995). Breakdown points for designed experiments. J. Statist. Plann. Inference 45 413-427. · Zbl 0827.62066 · doi:10.1016/0378-3758(94)00086-B
[19]	M üller, Ch. H. (1997). Robust Planning and Analysis of Experiments. Lecture Notes in Statist. 124. Springer, New York. · Zbl 0874.62079
[20]	Resnick, S. I. (1987). Extreme Values, Regular Variation, and Point Processes. Springer, New York. · Zbl 0633.60001
[21]	Rousseeuw, P. J. and Leroy, A. M. (1987). Robust Regression and Outlier Detection. Wiley, New York. · Zbl 0711.62030
[22]	Ruppert, D. (1992). Computing S estimators for regression and multivariate location/dispersion. J. Comput. Graph. Statist. 1 253-270.