Abstract
This paper describes a rounding procedure to improve the efficiency of index selection. The procedure involves performing canonical variate analysis on the phenotypic and genotypic variances of a group of traits estimated from a progeny test experiment. The eigenvectors corresponding to the significant eigenvalues are used to transform the original traits into a set of independent variables. The selection index is then constructed based on the new set of variables. The efficiency of the new index is expected to be improved by rounding off the variables associated with the insignificant eigenvalues.