Showing papers on "Robustness (computer science) published in 1976"

Confidence Interval Robustness with Long-Tailed Symmetric Distributions

Abstract: A variety of 95-percent confidence interval procedures have been examined in some detail using Monte Carlo techniques. These estimators were tried on simulated samples of sizes 10 and 20 from a spectrum of distributions ranging from the Gaussian to the long-tailed Cauchy. The robustness of an estimator is measured by both the closeness of its level to the 5-percent goal (robustness of validity) and its expected length as compared to its competitors (robustness of efficiency). Results include some quite robust procedures including some of the point M-estimators from the Princeton Robustness Study.

Three steps towards robust regression

Abstract: The three most commonly used statistics, the arithmetic mean, variance, and the product-moment correlation, are most unfortunate choices when data are not strictly Gaussian. A new measure of correlation and a measure of scale are proposed which are substantially more robust than their least squares counterparts. An illustration shows how increased robustness can be obtained through the use of equal regression weights without severe loss in accuracy. The paper also shows how incorporating knowledge about the theoretical structure of the regression coefficients into their estimation can aid substantially in increasing their robustness.

Duplicate analysis in geochemical practice. Part II. Examination of proposed method and examples of its use

Abstract: Monte Carlo simulation has been applied to test the robustness of a method for estimating precision as a function of concentration. The effect of deviations from the basic assumptions underlying the method are shown to be generally fairly small. The causes of such departures can be identified when they occur with actual laboratory results. Methods of recording laboratory observations can cause an over-optimistic bias of precision estimates in some circumstances.

Random Injection Control of Multiprogramming in Virtual Memory

Abstract: We propose a new method for the control of a multiprogrammed virtual memory computer system. A mathematical model solved by decomposition permits us to justify that the method avoids thrashing. Simulation experiments are used to test the robustness of the predictions of the mathematical model when certain simplifying assumptions are relaxed and when a slightly simpler control technique based on the same principle is used. Comparisons are given with the case where an "optimal" control is used and with that with no control. We also provide a simulation evaluating the estimators used in an implementation of the control, as well as the responsiveness of the controlled system to transients in the workload.

Robust studentization of location estimates

Abstract: Summary The robustness of power and of level of various estimators for the one sample location problem is compared by Monte Carlo methods; an estimate that studentizes particularly well is singled out. Generalizations to more complicated models are discussed.

The uniform distribution in signal detection theory

Abstract: The robustness of signal detection theory (SDT) is investigated with respect to the form of the underlying distributions. Usually these distributions are taken to be normal; here an SDT model based on two overlapping uniform (rectangular) distributions is examined, for the Yes/No experiment and the rating-method experiment. In the Yes/No case the SDT measure (using uniform distributions) is found to be equivalent to a measure recently proposed by Hammerton & Altham (1971), and, from contingency-table considerations, it is a measure likely to give similar conclusions to the SDT measure using normal distributions. In the rating-method case it is surprising to find that the SDT model with uniform distributions yields non-significant goodness-of-fit statistics for many sets of data.

Minimax robust estimation of location and minimizing fisher information

Abstract: Asymptotic minimax robust estimation of a location parameter as introduced by P. J. Huber is considered. This problem is closely related to recent developments in robust detection theory and robust recursive filtering for linear systems. For a given minimax location estimation problem two distinct estimates which are solutions are determined by the probability distribution Fo which has minimum Fisher information over the associated distribution set. A Lagrange multiplier technique is given for calculating Fo for a broad class of distribution sets. Existence and uniqueness conditions for Fo are given.