Medidas do grau de precisão experimental em ensaios de competição de cultivares de milho

Abstract

The objective of this work was to evaluate the appropriateness of some statistics as experimental precision measures and the relationships among them. Grain yield data from 101 cultivar competition trials of corn (Zea mays L.), which were carried out in the Rio Grande do Sul state, Brazil, in the agricultural years 2002/2003, 2003/2004 and 2004/2005, were used. Relationship among 12 statistics, estimated from each trial, was studied and correlation, path and cluster analysis were calculated based on these statistics. Class limits established by the statistics selective accuracy, heritability, coefficient of determination and F test value for cultivar were adequate to estimate the degree of experimental precision in corn yield trials. According to these statistics, 89% of the experiments showed a high or very high degree of precision and 5% of experiments could be discarded due to insufficient experimental precision. The statistics selective accuracy, heritability, coefficient of determination and F test value for cultivar show a direct relationship among them and are more adequate than the coefficient of variation and the significant minimum difference among cultivars by Tukey's test, expressed in percentage of the mean, to evaluate the experimental precision of corn yield trials. © 2009 Embrapa Informação Tecnológica.