Resumo

OSTENGO, Santiago; CUENYA, María I.  e  BALZARINI, Mónica. Mixed models in the analysis of multienvironment comparative yield trials of a perennial crop. Rev. ind. agric. Tucumán [online]. 2013, vol.90, n.2, pp.53-57. ISSN 1851-3018.

Multienvironment trials in sugarcane (Saccharum spp.) breeding are conducted through different crop ages, since sugarcane is a perennial crop. The temporal correlation among data recorded for the same plot across different years, as well as the spatial correlations among data from neighboring plots and the condition of heteroscedasticity among residual variances for each environment, can be approached in the context of mixed linear models (MLM). This work aims to compare the performance of sugarcane genotypes in a multienvironment trial under the MLM framework, which contemplates that there may be temporal and spatial correlations among yield data and that the heteroscedasticity condition may apply. Twenty sugarcane genotypes were evaluated in a multienvironmental trial through different crop ages. MLM that take into account correlations (temporal and/or spatial) and heterocedasticity in the structure of (co) variances of errors were adjusted for tons of cane per hectare (TCH) and sucrose content at the beginning of harvest (SC %). For TCH, the best model was the threedimensional first order autoregressive model, with temporal and spatial correlations and heteroscedasticity across locations. In the case of SC %, the heterocedastic first order autoregressive model, which includes only the temporal correlation, was the best. Both models provided more precise estimates (lower standard error) of the expected yield and changes in the relative performance of the compared genotypes.

Palavras-chave : Multienvironment trials; Spatial correlation; Temporal correlation; Saccharum spp..

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )