Rank Distance TO Choose Discriminators for Two Groups.
Rept. for Jul-Aug 71,
SCHOOL OF AEROSPACE MEDICINE BROOKS AFB TEX
Pagination or Media Count:
The rank distance method of classifying suggested by Zimmerman is defined for equisized training samples from two groups. A method for accepting or rejecting a proposed discriminator variable is described based on the number of misclassifications of the training set. To aid in implementing this choice, the distribution of the number of misclassifications under the null hypothesis is computed and presented for sample sizes up to twelve from each of the two groups. Some general characteristics of this distribution as a function of sample size are noted. Author
- Statistics and Probability