DIRECTION AND COLLINEARITY STATISTICS IN DISCRIMINANT ANALYSIS.
Abstract
Discrimination among several groups can be achieved with the help of a single optimum linear combination of correlated characters, if and only if all other linear combinations uncorrelated with it have no mean differences among the groups; and so, one is interested often in testing the goodness of fit of a hypothetical linear function for discrimination. The hypothetical function may not be good enough either (a) because the group means are not collinear, or (b) because the direction of the proposed function does not agree with the direction of the line on which the group means lie. Test statistics for these two aspects, (a) and (b), had been derived by Bartlett. In this paper, the distributions of these test statistics are derived by a simple analytical method instead of the geometrical devices used by Bartlett. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 20, 1969
- Accession Number
- AD0694440
Entities
People
- A. M. Kshirsagar
Organizations
- Southern Methodist University