ROBUST MODIFICATION OF GENOTYPE -BY -ENVIRONMENTS INTERACTION MODEL BY MONTE-CARLO SIMULATION

Abstract

Genotype Main Effects and Genotype -by -Environments Interaction (GGE) model is one of the frequently used models to capture and analyze Genotype- by-environment Interaction (GEI). The primary concern of most plant breeders and biometricians is to accurately model and analyze GEI, However. This could not be achieved in the GGE model as the model works on singular value decomposition (SVD), a method severely vulnerable to outlying observations. By a Monte Carlo simulation, this study modified the classical GGE model using three (3) robust SVD/PCA methods and obtained three (3) candidates GGE models namely: H-GGE, G-GGE and L-GGE. A simulated GGE multi-environment data was contaminated using pure shift scheme at various levels of generated outliers (2%,5%,10%,15%,20%,25% and 30%) to test and compare the performance of the models. The results revealed the vulnerability of the classical GGE model and further demonstrated robust performance of the modified models at the levels of the outliers used. The models were successfully tested on real multi-environment trials data involving twelve (12) genotypes of wheat grown in nine (9) environments obtained from Lake Chad Research Institute Maiduguri, Nigeria. We recommend to biometricians and plant breeders the use of the modified models for the robust analysis and interpretations of multi-environments data.