Adaptability and stability based on nonparametric regression in coffee genotypes

Abstract

The objective of this work was to evaluate a methodology of phenotypic adaptability and stability analyses of coffee genotypes based on nonparametric regression. The technique used differs from other techniques because it reduces the influence of extreme points resulting from the presence of genotypes whose answers to a certain environment are too different on the estimation of the adaptability parameter. Data from an experiment studying the average yield of 40 coffee (Coffea canephora) genotypes in a randomized block design with six replicates were used to evaluate the method. The genotypes were evaluated along five years (1996, 1998, 1999, 2000 and 2001) in two locations (Sooretama and Marilândia, ES, Brazil), in a total of ten environments. The methodology proposed proved adequate and efficient, since it eliminates the disproportionate effects induced by the presence of extreme points and avoids misleading recommendations of genotypes in terms of adaptability.