1U. Zannah 2A. Okolo 2D. Jibasen 2A.A. Akinrefon
1Department of Mathematics and Computer Science, Kashim Ibrahim University.
2Department of Statistics Modibbo Adama University Yola, Nigeria
*Corresponding author’s Email: umarzannah2@gmail.com, doi.org/10.55639/607.020100104
ABSTRACT
The use of factor analysis on adaptability and environmental stratification of genotypes and other agronomic and environmental studies has increased considerably over the years. Various methods were proposed on the analysis of adaptability and stratification of genotypes. One of such methods is the FGGE, that works on classical factor analysis in the matrix of genotypic effects (G) added to the effects of the genotype by environment interaction (GE), both fixed, obtained via the ordinary least squares method in joint analysis. The use of factor analysis obtained via the ordinary least squares method is vulnerable to outlying observations however small. To handle this limitation, this study proposed to modify the method using robust factor analysis on G+GE matrix. This proposed method is designated as M-FGGE. The G+GE matrix is obtained from simulated multi-environment trials data of 100 genotypes and 8 environments. Various levels of outliers (10%,20%,30% ,40%and 50%) were simulated and the G+GE matrix was contaminated using scattered environment contamination scheme. To assess and compare the performance of the modified method (M-FGGE) with its classical counterpart (FGGE) in handling outliers, the methods were tested on both contaminated and uncontaminated G+GE matrix. The eigenvalues, accumulated percentages of variance and communalities were obtained and compared. The results indicated that the modified method (M-FGGE) outperformed the existing method (FGGE). The study recommends to plant breeders and other researchers the use of M-FGGE as it offers robust application.
KEYWORDS
Robust factor
Analysis,
MCD,
G+GE matrix.
.

