Standardisation of techniques of clustering genotypes using mahalanobis D2 and wilks ^ criterion
Loading...
Files
Date
1986
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Department of Statistics, College of Veterinary and Animal Sciences, Mannuthy
Abstract
Two major drawbacks of Tocher's method of clustering genotypes using Mahalanobls D were pointed out and animprovement over Tocher's method was suggested. The cluster configuration obtained by these two methods were compared with those obtained by canonical analysis method.
A new computer oriented iterative algorithm for clustering using Mahalanobis D values was proposed.
A procedure for formation of clusters statistically, using O V Mahalanobls Er was suggested to form maximum nonsignificant
subsets of genotypes.
A new measure of dissimilarity which does not require any assumption on distribution of the population, viz., the determinant of the pairwise scatter matrix was proposed in the study.
Minimum /W/ criterion of Friedman and Rubin (1967) was also used for clustering. The clustering obtained by the new iterative algorithm using either Mahalanobls or determinant of pairwise scatter matrix or both could be used as the initial solution for it A graphical method for determining the optimum number of clusters was suggested The different methods were illustrated in two sets of data
Description
PG
Keywords
null