The selection of genetic materials in plant breeding programs. I. The discriminative power of the ranking

Abstract

In some phase of the plant improvement process, a large number of materials must be evaluated tor the selection of the best ones. In this case the experimental designs more frequently used in breeding programs are the square lattice designs, and only the best treatments based on the yield or another criterion are selected according to some selection criterion. Trying to get informations about the effect on the discriminative power of number of replications, coefficient of variation and discrepancy of the treatments in relation to the mean, 1400 experiments in a 5 x 5 square lattice design were simulated, with treatment effects ranging between + 25% and - 25% of the mean. It was verified that the ranking discriminatory power of the treatments are increased as the coefficient of variation gets lower and the number of replications increases. For the case of 12 replications and C.V. = 15% a discriminative power larger than 75% was observed when the difference between the selected material and the mean was of 25% of magnitude, and was less than 50% when these differences were situated between 10% and 7%.