Representations of the Space of Distributions in Robust Estimation of Location.
Abstract
In many situations it is useful to have a low-dimensional representation of the space of distributions. In this report, one, two and three dimensional representations are given which are of particular relevance to the study of robust estimation of location based on rank estimators. The distances are defined as functions of the asymptotic relative efficiency of the most efficient rank estimator for one distribution when used on data from another distribution. Values of these distance functions are computed for a large number of pairs of distributions and multidimensional scaling is used to find the low-dimensional representations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1981
- Accession Number
- ADA100562
Entities
People
- Brian L. Joiner
- David L. Hall
Organizations
- University of Wisconsin–Madison