The objective of this work was to validate a statistical methodology for estimating the reliability or normalized response (RNi) and stability of four elite rice genotypes, compared with the control IDIAP 145-05. We used the database from the rice breeding program IDIAP, from experiments conducted from 2009 to 2011, in 31 environments under dryland conditions. The results of the study allowed us to verify that new genotypes significantly exceeded (P<0.05) in grain yield the control used. The four genotypes IDIAP FL 106-11, 137-11 FL IDIAP, IDIAP FL 155 and FL 156 IDIAP had a reliability average of 0.79, 0.75, 0.75 and 0.74, respectively. The normalized probability IDIAP FL 106-11 represents a differential response greater than zero with respect to IDIAP 145-05, in 8 of 10 cases. It also confirmed the reliability is related to the stability parameters based on regression models (bi and S2di). The AMMI multivariate model, considering the PCA1, identified the genotype IDIAP FL 156, as the most stable. GGE Biplot model, based on the PCA2, found that the genotype IDIAP FL 155, had greater stability. The study illustrates that the reliability or standardized response may be useful for more precise recommendations for the use of new genotypes commercially.
Keywords: genotype by environment interaction, reliability or normalized response, phenotypic stability, rice breeding.