AMMI Model for Yield Estimation in Multi-Environment Trials: A Comparison to BLUP

Abstract

Maximum information from the Multi-environment trials (MET) can be reached by seeking the best estimator of each genotype's mean yield in a given environment. AMMI (additive main-effects and multiplicative interaction) is popular for analyzing MET data with fixed effect. When the environment included in MET is the sample of large environment, then environment effects regarded as random may be preferable, so the model is called mixed model. The assessment of it may be viewed as a problem of prediction rather than estimation. The prediction of the outcome of random variables is commonly done by Best Linear Unbiased Prediction (BLUP).Both methods are compared using the experimental rice data set from the Indonesian Rice Consortium's research which aims to evaluate the phenotypic performance of rice (Oryza sativa). Applying postdictive success method resultedAMMI10 as the best model, and its Root Mean Square Error Prediction is smaller than BLUP. AMMI was found to outperform BLUP in this rice dataset.