Bayesian analysis for genotype x environment interactions and the GGE-biplot assessment: Evaluation of balanced classifications with missing values
In multiple environmental trials (METs) most of the data, balanced or unbalanced, are normally tested over a wide range of environments (locations, years, growing seasons, etc.) and the basic statistical method used to obtain reliable statistical information. A case study is presented here to demonstrate the usefulness of Bayesian approach in genotype-by environment data analysis, in comparison with frequentist approach and GGE biplot assessment classification with missing value. Particular emphasis was given to Bayesian application that exploits pedigree information and to the analysis of GEI data for estimation of heritability, genetic gain and means prediction. A Markov Chain Monte Carlo (MCMC) method has been considered to perform Bayesian inference using R2WinBUGS. The study recently done in sorghum variety trials show investigation can be applied for multi environmental trial data. Results shows that the Bayesian estimation of variance components was accurate compared to the frequentist. The two principal components in GGEbiplot analysis were significant, explaining 95.13% (85.17% PC1 and 9.9.% PC2) for frequentist approach and explaining 97.36% (84.06% PC1 and 13.3% PC2) for Bayesian approach of interaction variation. Bayesian analysis indicates GGE-biplot gave the best results in contributing to the GEI. Bayesian approach for analysis GEI data is highly suitable with missing values.
Int J Appl Sci Biotechnol, Vol 3(2): 210-217