Scholion.ai

Alternatives to the Median Absolute Deviation

Peter J. Rousseeuw, Christophe Croux

Journal of the American Statistical Association · 1993 · Volume 88 · Issue 424

DOI: 10.1080/01621459.1993.10476408

Abstract

In robust estimation one frequently needs an initial or auxiliary estimate of scale. For this one usually takes the median absolute deviation MAD n = 1.4826 med, {| x i − med j x j |}, because it has a simple explicit formula, needs little computation time, and is very robust as witnessed by its bounded influence function and its 50% breakdown point. But there is still room for improvement in two areas: the fact that MAD n is aimed at symmetric distributions and its low (37%) Gaussian efficiency. In this article we set out to construct explicit and 50% breakdown scale estimators that are more efficient. We consider the estimator S n = 1.1926 med, {med j | x i − x j |} and the estimator Q n given by the .25 quantile of the distances {| x i − x j |; i < j }. Note that S n and Q n do not need any location estimate. Both S n and Q n can be computed using O ( n log n ) time and O ( n ) storage. The Gaussian efficiency of S n is 58%, whereas Q n attains 82%. We study S n and Q n by means of their influence functions, their bias curves (for implosion as well as explosion), and their finite-sample performance. Their behavior is also compared at non-Gaussian models, including the negative exponential model where S n has a lower gross-error sensitivity than the MAD.

Enable JavaScript for citation tools and additional paper interactions.