Interação genótipos x ambientes e variação espacial em experimentos de avaliação de genótipos de milho no Brasil Central

Abstract

For a promising selection of productive genotypes, accurate statistical procedures should be used to improve most precision in the comparision of means. This greater precision and accuracy can be obtained by the use of different statistical tools such as the study of adaptability and stability of genotypes in evaluation and also by the detection of spatial dependence, which may result in greater spatial variability of the study variables. The first chapter had as objectives: 1. To estimate parameters of adaptability and stability of productivity in differents environments using five estimation methods. 2. To compare five adaptability and stability methods regarding the discriminatory capacity to recommendation of maize genotypes to Central Brazil region. To the first chapter, adaptability and stability analyzes with different statistical methods were performed, considering data from 36 maize genotypes grown in nine different cultivation regions of the Central Brazil region. We used the methods Eberhart e Russel (1966), Lin e Binns (1988) with decomposition CRUZ; REGAZZI; CARNEIRO, 2012), additive Main effect and Multiplicative Interaction (AMMI-Biplot), genotype+genotype interaction (GGE-Biplot) and harmonic means of the relative performance of genetic values by REML/BLUP (MHPRVG). The methods have similarities in ordering genotypes, but they are different in relation to the precision and information provided about GxE interaction. GGE-Biplot method retains a greater proportion of the total square sum and may report tha the simple interaction was predominant when compared to AMMI-Biplot method. Lin e Binns (1988) with decomposition was able to provide genotypes productive, adapted and stable, under the simple type interaction. The MHPRVG and GGE-Biplot methods should be used together to select most promising genotypes. Sint. 10771 and Sint. 10697 were recommended for cultivation in the Central Brazil region due to theis adaptability, stability and productivity. In the second chapter, the objectives were: 1. To compare parameters o experimental precision of traditional variance analysis with the analysis of variance by autoregressive model in 14 maize experiments with nitrogen fertilization condition (ideal) and stress condition (low); and, 2. To verify if there is difference with respect to the correlated erroes between experiments with nitrogen fertilization condition (ideal) and stress condition (low). The analysis of variance by autoregressive model provided values of parameters of experimental precision similar to those expressed by the analysis of traditional variance and there was no difference in relation to the errors correlated in experiments with nitrogen fertilization condition (ideal) and stress condition (low).