A Robust Class of Tests and Estimates for Multivariate Location

Abstract
Hotelling's T 2 test and the sample mean vector for multivariate location suffer from the same lack of robustness as the univariate t test and sample mean. We propose a class of tests and estimates based on a vector of rank statistics and show that they are easy to compute, have good robustness properties, and are usually more efficient than the classical procedures.

This publication has 0 references indexed in Scilit: