A Comparison of Some Robust Procedures for Estimating a Linear Discriminant Function,
Abstract
A number of methods have been suggested for robustly estimating a linear discriminant function. These include substitution of robust estimates for the mean and covariance matrix and methods which choose a projection to maximize a robust measure of separation. This paper presents results of Monte Carlo simulations comparing some of these methods along with various modifications to see whether relatively simple methods works as well as complicated ones.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADP007119
Entities
People
- Hongzhe Li
Organizations
- University of Montana