The objectives of this work were to establish repeatability estimates for agro-industrial characteristics for twenty genotypes of sugar cane, to indicate the determination coefficient of each character and the prediction of truthful value of each clone. The design consisted of randomized blocks, with five replications. The analysis of variance with two factors of variation (cuts and genotypes) and the analysis of the main components were used to estimate the repeatability coefficient <img border=0 width=19 height=21 id="_x0000_i1028" src="../../../../img/revistas/pab/v39n4/20437s1.gif" align=absmiddle>, initiating with three cuts. Estimates of repeatability were obtained over 0.5 for fiber and ton of cane per hectare, in both methods, with confidence over 84% by the method of main components. Characteristics lower than 0.5 and determination lower than 74% need a greater number of measures. The methods of main components and analysis of variance evidenced, in five cuts, determination over 80% for fiber, percentage of pol (sucrose) in the juice of cane, tons of cane per hectare and tons of pol in the juice of cane per hectare, though the first had been most efficient. Considering tons of cane per hectare and tons of pol in the juice of cane per hectare, the clones RB9371, RB9350 and RB9364 are the best.
Saccharum; clones; selection; analysis of variance; principal components
