Multi-environment trials data analysis: An efficient biplot analysis approach
Abstract The analysis of multi-environment trials (MET) data has a long history in plant breeding and agricultural research, with the earliest approaches being based on ANOVA methods. ANOVA-based biplot analysis has been used for a long time in analyzing MET data, and advances have been made employing different modeling approaches. This paper presents MET data analysis using mixed model approaches, and compares three methods of biplot analysis, namely genotype main effects plus genotype by environment interaction (GGE) analysis, factor analytic multiplicative mixed (FAMM) model analysis, and combined model analysis. Ten grain yield datasets from the national variety trial series conducted by the Ethiopian institute of agricultural research were used for this study. Our results revealed that spatial and FA model provide a significant improvement in analyzing MET data. This was demonstrated with evidence of heritability measure. We demonstrated that biplot analysis based on the approached of combined model analysis provides a substantial increase in the total percentage of genotype by environment (G×E) variance explained by the first two multiplicative components for both types of balanced and unbalanced datasets. Thus, by estimating the G×E mean values with the best linear unbiased predictions using spatial+FA (FAMM model analysis), and thereby conducting biplot analysis based on the combined model analysis, plant breeding and trial evaluation programs can have a more robust platform for evaluation of crop cultivars with greater confidence in discriminating superior cultivars across a range of environments.