-

In order to study variations among traits detecting suitable selection index, clustering of genotypes regarding their similarity, and experiment with 100 safflower genotypes was conducted at Research Farm of Seed and Plant Improvement Institute in 1998. The experimental design was a 10x10 simple lattice. The results of analysis of variance demonstrated that the differences among genotypes were highly significant (P<0.01) for all traits. High values of phenotypic and genotypic coefficients of variation were obtained for most traits, indicating high variability in the traits under study. The recorded da.ta
were subjected to principal component and factor analysis. The results showed that seven principal components and factors with eigen values more than one explained 80.9% of the total variability. Cluster analysis (Ward method) calssified the genotypes based on traits related to grain yield and geographical
origins in 6 and 13 clusters, respectively. The obtained results revealed that genotypes could be more effectively clustered when considering traits related to grain yield in comparison with geographical origin.