University of Washington - Department of Statistics
This talk is a personalized account of John Tukey's contributions to robust statistics, as well as a summary of the maturation of robustness theory and practice to date. I begin by fondly recalling the way in which Tukey and I became acquainted, how he gave me my start in Statistics at Princeton and Bell Laboratories, and the very stimulating research environment of the Mathematics and Statistics Research Center at Bell Laboratories in 1970's and 1980's. Then I will review the following aspects of Tukey's contributions to robustness: efficiency robustness, resistance, sensitivity curves, the bi-weight M-estimate score-like function, bi-efficiency and tri-efficiency, robust estimator selection as empirical science and the "Princeton" robustness study, Monte Carlo swindle techniques, robust smoothers and twicing, robust regression, "useable" robust methods, and the follow-on Princeton robustness meetings. Above all, Tukey was a powerful stimulus to the research of many other "robustniks", and to a greater or lessor degree inspired the fundamental subsequent work of Huber and Hampel, which I will briefly summarize. The ultimate payoff of years of research on robust statistical methods will be in their widespread use to quickly and easily provide good fits to the bulk of the data, detect multidimensional outliers and provide robust inference. I will close by briefly demonstrating use of the S-PLUS Robust Library to achieve these goals.