Genetic dissimilarity measures are commonly used by plant breeders to identify different genotypes to get desired segregant populations. This study was proposed to establish the relationship between different multivariate techniques to estimate divergence. The experiment was performed during the growing season of 2001, where twelve oat cultivars were tested for seven agronomic traits, using the random blocks experimental design with four replications. Euclidian and Mahalanobis distances showed low correlation (0.529) and when used to build dendrograms did not show similar clustering. The graphic techniques analysis through principal components and canonical variables also showed distinct spreading patterns. However, in spite of the observed discrepancies among the methodology analyzed, it was possible to recognize dissimilar genotypes with high average that can be used with large success probability in selected artificial hybridizations in oats.
choice of parents; principal components; canonical variables; cluster analyses
