The Annals of Statistics

Optimum robust testing in linear models

Christine Müller

Full-text: Open access

Abstract

Robust tests for linear models are derived via Wald-type tests that are based on asymptotically linear estimators. For a robustness criterion, the maximum asymptotic bias of the level of the test for distributions in a shrinking contamination neighborhood is used. By also regarding the asymptotic power of the test, admissible robust tests and most-efficient robust tests are derived. For the greatest efficiency, the determinant of the covariance matrix of the underlying estimator is minimized. Also, most-robust tests are derived. It is shown that at the classical $D$-optimal designs, the most-robust tests and the most-efficient robust tests have a very simple form. Moreover, the $D$-optimal designs provide the highest robustness and the highest efficiency under robustness constraints across all designs. So, $D$-optimal designs are also the optimal designs for robust testing. Two examples are considered for which the most-robust tests and the most-efficient robust tests are given.

Article information

Source
Ann. Statist., Volume 26, Number 3 (1998), 1126-1146.

Dates
First available in Project Euclid: 21 June 2002

Permanent link to this document
https://projecteuclid.org/euclid.aos/1024691091

Digital Object Identifier
doi:10.1214/aos/1024691091

Mathematical Reviews number (MathSciNet)
MR1635454

Zentralblatt MATH identifier
0929.62080

Subjects
Primary: 62F35: Robustness and adaptive procedures 62K05
Secondary: 62J05 62J10

Keywords
Linear model shrinking contamination robust tests asymptotically linear estimators bias of the level most robustness efficiency $D$-optimality optimal design

Citation

Müller, Christine. Optimum robust testing in linear models. Ann. Statist. 26 (1998), no. 3, 1126--1146. doi:10.1214/aos/1024691091. https://projecteuclid.org/euclid.aos/1024691091


Export citation

References

  • Akritas, M. G. (1991). Robust M estimation in the two-sample problem. J. Amer. Statist. Assoc. 86 201-204.
  • Bickel, P. J. (1981). Quelque aspects de la statistique robuste. ´Ecole d' ´Et´e de Probabilit´es de St. Flour. Lecture Notes in Math. 876 1-72. Springer, Berlin.
  • Bickel, P. J. (1984). Robust regression based on infinitesimal neighborhoods. Ann. Statist. 12 1349-1368.
  • Christensen, R. (1987). Plane Answers to Complex Questions-The Theory of Linear Models. Springer, New York.
  • Coakley, C. W. and Hettmansperger, T. P. (1992). Breakdown bounds and expected resistance. J. Nonparametr. Statist. 1 267-276.
  • Coakley, C. W. and Hettmansperger, T. P. (1994). The maximum resistance of tests. Austral. J. Statist. 36 225-233.
  • H´ajek, J. and Sid´ak, Z. (1967). Theory of Rank Tests. Academic Press, New York.
  • Hampel, F. R. (1978). Optimally bounding the gross-error-sensitivity and the influence of position in factor space. In Proceedings of the Amer. Statist. Assoc. Statistical Computing Section 59-64. ASA, Washington, DC.
  • Hampel, F. R., Ronchetti, E. M., Rousseeuw, P. J. and Stahel, W. A. (1986). Robust StatisticsThe Approach Based on Influence Functions. Wiley, New York.
  • Heritier, S. and Ronchetti, E. (1994). Robust bounded-influence tests in general parametric models. J. Amer. Statist. Assoc. 89 897-904.
  • Kiefer, J. (1974). General equivalence theory for optimum designs (approximate theory). Ann. Statist. 2 849-879.
  • Kiefer, J. and Wolfowitz, J. (1960). The equivalence of two extremum problems. Canad. J. Math. 12 363-366.
  • Krasker, W. S. (1980). Estimation in linear regression models with disparate data points. Econometrica 48 1333-1346.
  • Krasker, W. S. and Welsch, R. E. (1982). Efficient bounded-influence regression estimation. J. Amer. Statist. Assoc. 77 595-604.
  • Kurotschka, V. and M ¨uller, C. H. (1992). Optimum robust estimation of linear aspects in conditionally contaminated linear models. Ann. Statist. 20 331-350.
  • Markatou, M. and He, X. (1994). Bounded influence and high breakdown point testing procedures in linear models. J. Amer. Statist. Assoc. 89 543-549.
  • Markatou, M. and Hettmansperger, T. P. (1990). Robust bounded-influence tests in linear models. J. Amer. Statist. Assoc. 85 187-190.
  • Markatou, M. and Manos, G. (1996). Robust tests in nonlinear regression models. J. Statist. Plann. Inference 55 205-217.
  • Markatou, M., Stahel, W. A. and Ronchetti, E. (1991). Robust M-ty pe testing procedures for linear models. In Directions in Robust Statistics and Diagnostics I (W. Stahel and S. Weisberg, eds.) 201-220. Springer, New York.
  • M ¨uller, C. H. (1987). Optimale Versuchspl¨ane f ¨ur robuste Sch¨atzfunktionen in linearen Modellen. Ph.D. thesis. Freie Univ. Berlin. M ¨uller, C. H. (1992a). L1-estimation and testing in conditionally contaminated linear models. In L1-Statistical Analy sis and Related Methods (Y. Dodge, ed.) 69-76. North-Holland, Amsterdam. M ¨uller, C. H. (1992b). Robust estimation with minimum bias and A-optimal designs. In PROBASTAT'91, Proceedings of the International Conference on Probability and Mathematical Statistics, Bratislava 1991 (A. P´azman and J. Volaufov´a, eds.) 109-115. M ¨uller, C. H. (1992c). Robust testing in conditionally contaminated linear models. Preprint No. A-92-12, Fachbereich Mathematik und Informatik, Freie Univ. Berlin. M ¨uller, C. H. (1994a). Optimal designs for robust estimation in conditionally contaminated linear models. J. Statist. Plann. Inference 38 125-140. M ¨uller, C. H. (1994b). One-step-M-estimators in conditionally contaminated linear models. Statist. Decisions 12 331-342.
  • M ¨uller, C. H. (1995). Outlier robust inference for planned experiments. Habilitation thesis. Freie Univ. Berlin.
  • M ¨uller, C. H. (1997). Robust Planning and Analy sis of Experiments. Lecture Notes in Statist. 124. Springer, New York.
  • P´azman, A. (1986). Foundations of Optimum Experimental Design. Reidel, Dordrecht.
  • Pukelsheim, F. (1993). Optimal Design of Experiments. Wiley, New York.
  • Rieder, H. (1985). Robust estimation of functionals. Technical Report, Univ. Bay reuth.
  • Rieder, H. (1987). Robust regression estimators and their least favorable contamination curves. Statist. Decisions 5 307-336.
  • Rieder, H. (1994). Robust Asy mptotic Statistics. Springer, New York.
  • Ronchetti, E. (1982). Robust testing in linear models: the infinitesimal approach. Ph.D. thesis, Swiss Federal Institute of Technology, Z ¨urich.
  • Ronchetti, E. and Rousseeuw, P. J. (1985). Change-of-variance sensitivities in regression analysis. Z. Wahrsch. Verw. Gebiete 68 503-519. Silvapulle, M. J. (1992a). Robust Wald-ty pe tests of one-sided hy pothesis in the linear model. J. Amer. Statist. Assoc. 87 156-161. Silvapulle, M. J. (1992b). Robust tests of inequality constraints and one-sided hy potheses in the linear model. Biometrika 79 621-630.
  • Staudte, R. G. and Sheather, S. J. (1990). Robust Estimation and Testing. Wiley, New York.
  • Wald, A. (1943). Tests of statistical hy pothesis concerning several parameters when the number of observations is large. Trans. Amer. Math. Soc. 3 426-482.