×

On the explosion rate of maximum-bias functions. (English) Zbl 0941.62035

Summary: The maxbias function \(B_T(\varepsilon)\) contains much information about the robustness properties of the estimate \(T\). This function satisfies \(B_T(O)=0\) and \(B_T(\varepsilon) <\infty\) for all \(0<\varepsilon <\varepsilon^*\), where \(\varepsilon^*\) is the breakdown point of \(T\). F. R. Hampel [J. Am. Stat. Assoc. 69, 383-393 (1974; Zbl 0305.62031)] pioneered the study of the limiting behaviour of \(B_T(\varepsilon)\) where \(\varepsilon\to 0\). He computed and optimized the rate \(\gamma\) at which \(B_T (\varepsilon)\) approaches 0 when \(\varepsilon\to 0\). This rate is now called the contamination sensitivity of \(T\), and constitutes one of the cornerstones of the theory of robustness.
We show that much can also be learned from the study of the limiting behaviour of \(B_T(\varepsilon)\) when \(\varepsilon\to \varepsilon^*\). A new robustness measure, called the relative explosion rate, can be obtained by studying the limiting relative maxbias behaviour of two extimates when \(\varepsilon\) approaches their common breakdown point \(\varepsilon^*\). Like the contamination sensitivity, the relative explosion rate can be readily derived from the estimate’s score function. General formulae are given for \(M\)-estimates of scale and \(S\)-, \(MM\)- and \(\tau\)-estimates of regression. We also show that the maxbias behaviour for large \(\varepsilon\) is largely determined by the curvature of the estimate’s score function near zero. This motivates our definition and study of the local order of a score function.

MSC:

62F35 Robustness and adaptive procedures (parametric inference)
62F10 Point estimation

Citations:

Zbl 0305.62031
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Berrendero, On the maxbias curve of residual admissible robust regression estimates (1995)
[2] Croux, Efficient high-breakdown M-estimators of scale, Statist. Probab. Leu. 19 pp 371– (1994) · Zbl 0791.62034
[3] Donoho, A Festchrift for Erich A. Lehmann pp 157– (1983)
[4] Hampel, The influence curve and its role in robust estimation, J. Amer. Statist. Assoc. 69 pp 383– (1974) · Zbl 0305.62031
[5] He, Lower bounds for contamination bias: Globally minimax versus locally linear estimation, Ann. Statist. 21 pp 314– (1993) · Zbl 0770.62023
[6] Huber, Robust estimation of a location parameter, Ann. Math. Statist. 35 pp 73– (1964) · Zbl 0136.39805
[7] Huber, Robust Statistics. (1981)
[8] Martin, Asymptotically min-max bias robust M-estimates of scale for positive random variables, J. Amer. Statist. Assoc. 17 pp 494– (1989)
[9] Martin, Min-max bias robust regression, Ann. Statist. 17 pp 1608– (1989) · Zbl 0713.62068
[10] Rousseeuw, The bias of k-step M-estimators, Statist. Probab. Lett. 20 pp 411– (1994) · Zbl 0801.62036
[11] Rousseeuw, Robust Regression and Outlier Detection. (1987)
[12] Rousseeuw, Robust and Nonlinear Time Series Analysis pp 256– (1984)
[13] Rupport, Computing S-estimators for regression and multivariate location/dispersion, J. Comput. Graph. Statist. 1 pp 253– (1992)
[14] Yohai, High breakdown point and high efficiency robust estimates for regression, Ann. Statist. 15 pp 642– (1987) · Zbl 0624.62037
[15] Yohai, High breakdown point estimates of regression by means of the minimization of an efficient scale, J. Amer. Statist. Assoc. 83 pp 406– (1988) · Zbl 0648.62036
[16] Yohai, A minimax bias property of the least {\(\alpha\)}-quantile estimates, Ann. Statist. 20 pp 1875– (1993) · Zbl 0797.62027
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.