AMMI Analysis of Four‐Way Genotype × Location × Management × Year Data from a Wheat Trial in Poland

• Published in: Crop science 2016-09, Vol.56 (5), p.2157-2164
• Main Authors: Paderewski, Jakub, Gauch, Hugh G., Mądry, Wieslaw, Gacek, Edward
• Format: Article
• Language: English
• ISSN: 0011-183X; 1435-0653
• DOI: 10.2135/cropsci2015.03.0152

Grain yield data of winter wheat (Triticum aestivum L.) trials in Poland had a four‐way factorial design of 24 genotypes by 20 locations by two managements by 3 yr. The experimental design had genotype–management strip plots with two replications for genotypes, with somewhat more genotypes than the 24 having no missing data. The research objectives were to extend additive main effects and multiplicative interactions (AMMI) analysis from two‐way to higher‐way datasets to reduce spurious complexity originating from noise, delineate wheat mega‐environments in Poland, and make genotype recommendations within each mega‐environment. Statistical analysis began with adjusting the yield estimates using the strip‐plot experimental design and then combining the results in a genotype × location × management × year (GLMY) table. This table was analyzed by a four‐way ANOVA mixed model. Next the GLMY dataset was reorganized into a two‐way classification, namely a genotype × environment (G × E) dataset, where the 120 environments were defined as combinations of location, management, and year. This two‐way dataset was analyzed by AMMI, with practical limitations of working with only a few mega‐environments focusing interest on the AMMI1 member of this model family. The first principal component had an evident geographical interpretation, contrasting northeast Poland (colder climate) and southwest Poland (warmer climate). Suitable genotypes were recommended within each of these two mega‐environments. The methodological significance of this paper is the extension of AMMI analysis from the customary two‐way G × E datasets to higher‐way datasets, such as the present four‐way GLMY dataset.