, rabiei@guilan.ac.ir
Abstract: (26093 Views)
Discrimination function analysis is a method of multivariate analysis that can be used for determination of validity in cluster analysis. In this study, Fisher’s linear discrimination function analysis was used to evaluate the results from different methods of cluster analysis (i.e. different distance criteria, different cluster procedures, standardized and un-standardized data). Furthermore, Hotelling T2, CCC plot and multivariate analysis of variance were used to support the results. To achieve the goals, 8 rapeseed genotypes were planted in a randomized complete block design with three replications in Rice Research Institute of Iran, Rasht, durin 2005-2006, and 14 characteristics were measured. Analysis of variance based on the randomized complete block design showed significant differences between genotypes for all the studied traits. Comparison of means between genotypes indicated that the genotype Hyola401 for grain yield and most of the measured characteristics was better than the other genotypes. Evaluation of phenotypic and genotypic coefficient of variations showed that most of the traits had high variability in the population. Discrimination function analysis showed that the Euclidean distance criterion was better than others and a desirable clustering was obtained by this criterion. Also, all of the data standardization methods produced similar clusters and were better than un-standardized data. Based on evaluation of dendrograms derived from different methods of cluster analysis determined that the UPGMA, complete linkage and Ward’s minimum variance methods were better than the other methods, and grouped the genotypes into three clusters. Fisher’s linear discrimination function analysis showed that UPGMA and Ward's minimum variance methods with clustering validity of 87.5 percent, was more suitable than other cluster analysis methods however, discrimination analysis grouped genotypes into two clusters. Tests of Hotelling T2, CCC plot and multivariate analysis of variance supported the results from the discrimination function analysis. It seems that the UPGMA and Ward's minimum variance procedures based on Euclidean distance criterion of standardized data function better in grouping genotypes, yet, the use of discrimination function analysis is recommended to confirm the results and determine the actual groups.
Type of Study:
Research |
Subject:
Ggeneral Received: 2010/05/12 | Published: 2009/04/15