CHOICE OF AXES FOR PROJECTION OF DATA ON A CRT USING MULTIVARIATE ANALYSIS OF VARIANCE,
Abstract
Display on a CRT of multi-dimensional vector data requires that the points be projected onto a two-dimensional subspace. If the purpose of the display is to determine how a given set of measurements distinguishes among a number of known classes of objects, then we can use discriminant techniques and multivariate analysis of variance to help choose good subspaces in which to view the data. This report describes the analysis, noting some ways in which one and two way analysis can be used to generate effective displays. Numerical methods for solving the conditioned eigenvalue problem are surveyed, and various consequences are developed of one of these algorithms, which allows the analysis to be performed on systems with very many measurements (100 - 1000). (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- May 15, 1968
- Accession Number
- AD0678105
Entities
People
- Robert M. Colomb