The experiment was carried out to evaluate the coefficient of repeatability of some agronomic characteristics in Panicum maximum hybrids. Ten plots of sexual plants of this species were randomly distributed among 230 plots of apomitic accessions. After natural hybridization, seeds of each sexual plant constituted a half-sib family. Thirty plants of each female progenitor were evaluated in a randomized blocks design, with five plants per plot and six replications. Based on five evaluation cuts, the coefficient of repeatability was estimated by the methods analysis of variance, principal components (correlation and covariance matrices) and structural analyses. The coefficient of repeatability (r) estimated by the different methods, for all the characteristics evaluated, was high and varied from 0.51 to 0.86. The estimates of the coefficient of determination were also high. It was verified that the estimates of the repeatability coefficient for the four characteristics evaluated by the analyses of variance method were almost always smaller than the estimates obtained by the other methods. By the principal component method (based on the covariance matrix), the estimates were always the highest. Considering a level of 80 or 90% as satisfactory for the confidence in the decision of the relative superiority of the hybrids based on green matter yield, these five evaluation cuts would be sufficient for the choice of the best plant. The exclusion of the first two cuts will increase the repeatability and determination coefficients.
analysis of variance; dry matter production; principal components; structural analysis
