The Selection of Effective Attributes for Deciding Between Hypotheses Using Linear Discriminant Functions.
Abstract
Let X be a multidimensional random variable, whose components are called attributes, with mean mu sub i and covariance matrix Sigma sub i under H sub i, i=1, 2. The problem discussed is related to problems in regression theory. Some simple examples illustrate that if Sigma sub 1 is in some sense much smaller than Sigma sub 2, there is a premium on adjoining additional attributes for which the variance under H sub 1 is relatively small.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 15, 1970
- Accession Number
- AD0716956
Entities
People
- Herman Chernoff
Organizations
- Stanford University