Misclassification Rates of Likelihood and Predictive Discriminant Functions for Small Samples
Abstract
Likelihood and predictive discriminant function misclassification rates are compared when training sets are small (i.e., less than 20). The theoretical foundations for the linear, quadratic, and predictive discriminant functions are described. Simulations are used to compare the classification capability of each discriminant function while varying the variance of the underlying distributions. A multivariate case is also analyzed.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 01, 1992
- Accession Number
- ADA253640
Entities
People
- Don Waagen
Organizations
- Naval Command, Control and Ocean Surveillance Center