In the breeding of perennial plant species, the coefficient of repeatability is an important parameter because it allows effective early selection of superior genotypes, thus saving time and cost. The objective of this study was to determine the minimum numbers of years required for observation to identify superior genotypes of coconut with high degree of accuracy. Four hybrid varieties of coconut were evaluated in a randomized complete block with two replications in each of four years (2009 to 2012) in Benin City, Edo State, Nigeria. Repeatability coefficient was calculated using principal components based on correlation and covariance matrices, and the minimum numbers of years for observation to identify superior genotypes were estimated. The repeatability coefficient ranged from 0.63 to 0.90, while the coefficient of determination ranged from 0.87 to 0.97 for the two methods of principal component analysis for yield and its components. For each of the traits studied, the repeatability coefficient estimated by principal components based on correlation matrix was lower than the value obtained by the principal components based on the covariance matrix. Estimates of the number of years required for observation to identify superior genotypes based on correlation and covariance matrix of the principal component analysis ranged from 2 to 58 years at a pre-established reliability coefficient of 80 to 99%. Results from the simulation models showed that an increase in number of years of evaluation beyond four did not appreciably increase the precision for the selection of superior genotypes (R2> 85%) for each of the traits evaluated. Based on repeatability index, four year yield records are adequate for the identification of superior genotypes with a high degree of accuracy.

Keywords: Repeatability, hybrid varieties, principal components, reliability coefficient.