Regression and uni-multivariate methodologies for reduction of the replication number in experiments of the intermediary phase of a soybean breeding program. / Métodos de regressão e uni-multivariado para a redução do número de repetições em experimentos intermediários de um programa de melhoramento de soja.

AUTOR(ES)

Fernando Toledo Santos de Miranda

DATA DE PUBLICAÇÃO

2004

RESUMO

The intermediary phase of a soybean breeding program involves the evaluation of a large number of genotypes (about 100 lines) in several environments, becoming this a very expensive step. The utilization of statistical methods that allow a refined analysis of the genotype x environment (GxE) interaction, may generate gains in precision as a compensation to the expected increase in the GxE estimate and, thus, to permit the reduction of the replication number. Eberhart &Russell (1966) (ER) methodology utilizes linear regression to study G x E interaction; the AMMI methodology employs the analysis of variance to fit effects of genotypes and environments and singular values decomposition to fit only the GxE interaction. The objective of this research was to evaluate experiments with 72 lines in randomized block design with two replications subdivided in sets with common checks (BCCE); these experiments were carried out in four locations / sow dates during three agriculture years. The experiments were analyzed in BCCE and also in augmented blocks (BA) by considering only one replication taking at random. As auxiliary tools were used ER and AMMI methodologies. The joint analysis of the 12 environments through the AMMI methodology was used as pattern for comparisons through Spearman correlation (rs). In relation to this pattern, the mean of rs in the three years was estimated in: 54% for means of BA experiments, 64% for BCCE experiments, 65% for BA experiments analyzed with ER, 69% for BCCE experiments analyzed with ER methodology, 73% for BA experiments analyzed with AMMI methodology and, 74% for BCCE experiments analyzed with AMMI methodology. The results indicated that the application of auxiliary methods for understanding GxE interaction were able to increase the rs with the pattern, opening the possibility for reducing the replication number in the experiments of the intermediary steps of soybean breeding programs. In conclusion, it was verified the possibility to reduce from two (BCCE) by one (BA) replication by using the following auxiliary methods of analysis: a) AMMI method instead of means of two replications as the unique selection criterion; b) ER method (in two of the three evaluated years) instead of means of two replications as the unique selection criterion; c) AMMI method (in two of the three evaluated years) instead of selection as based on ER method. The reduction from two to one replication makes possible to lower reasonably the experimental costs during the intermediary step of breeding programs.

ASSUNTO(S)

breeding imbreed lines análise de variância soybean delineamento experimental soja interaction genotipe environment melhoramento genético vegetal linhagens vegetais analysis of variance experimental designes

Documentos Relacionados

• Adaptabilidade e estabilidade de linhagens experimentais de soja selecionadas para caracteres agronômicos através de método uni-multivariado
• Otimização do número de progênies e de repetições em experimentos de melhoramento de plantas
• Modelos de regressão multivariada
• Método de Papadakis e número de repetições em experimentos de soja
• "Analysis on a multivariate null-intercept errors-in-variables regression model"