Abstract :
In final stages of plant breeding programs, a large number of new improved genotypes are tested over a wide range of test environments and the underlying statistics used to model this system may be rather complicated. Usually, the presence of the genotype × environment (GE) interaction effect complicates the selection of the most favorable genotypes for a target test environment. There are several statistical methods available to analyze results of multi-environment trials including a range of univariate and multivariate procedures. Univariate methods have inadequate capacity to fully explain the GE interaction structure because they attempt to define the GE interaction by one or two parameters but the multiplicative GE interaction is far too complex to be summarized by only some limited parameters. In contrast, multivariate statistical methods explore multi-directionality aspects of the GE interaction and try to extract more information. The most common multivariate statistical methods are cluster analysis (CA), principal components analysis (PCA), principal coordinates analysis (PCOA), factor analysis (FA), the additive main effect and multiplicative interaction (AMMI), shifted multiplicative model (SHMM), site regression biplot (GGE). This paper reviews these multivariate statistical methods for analyzing a multi-environment trial dataset. Several AMMI stability parameters were discussed and three of these important models (AMMI, GGE and SHMM) are compared.
