The objective of this study was to estimate the repeatability coefficient in sweet orange and the number of repeated measures that should be performed for an efficient selection of genotypes. From March to June 1999, chips were collected from producing sweet orange seedlings, in nine municipalities of Acre, Brazil, and were grafted on 'Rangpur' lime rootstock, resulting in 54 clones, which were evaluated together with the cultivar Aquiri, recommended for the state. The experiment was carried out in a randomized complete blocks design, with 55 treatments, three replicates and one plant per plot. Fruit number per tree, fruit production per tree and average fruit weight were evaluated in 2002, 2003, 2004, 2005, 2006 and 2008. Analysis of variance, analysis of principal components, and structural analysis were performed. The estimate of repeatability for average fruit weight showed regularity in the ranking of clones, from one cycle to another, and five cycles of assessments were necessary to predict the real value of clones, with accuracy of 90%, when principal components method (matrix of co-variance) was used. Although the number of fruits per plant and fruit yield per plant are influenced by the environment, eight and nine measurements, respectively, would achieve determination coefficients of 95%.
structural analysis; multivariate analysis; plant breeding; clone selection
