Repeatability and number of harvests required for selection in robusta coffee

Abstract

This study aimed to estimate the repeatability coefficient of the grain yield in Coffea canephora by three methods: to quantify the precision of the measurements; to predict the real value of an individual based on n evaluations; and to determine the number of phenotypic measures required in each plant to obtain an adequate precision level for an efficient discrimination of the genotypes. The coefficients of repeatability and determination were estimated based on four harvests of 80 genotypes. Highest estimates of the repeatability coefficient were obtained by the method of the principal components derived from the matrix of covariances, which expresses the correlation between each measurement pair. The prediction precision of the real individual value ranged from 65.32 to 81.59%, and remained practically unchanged from the sixth harvest on.