Outlying Observation in Stability Analysis of Genotype: AMMI vs Huehn Method

Abstract

Producing a quality crop that has superior characteristics is one of the important things in agronomy. An assessment is needed to see if the superiority is stable in various environmental conditions or various locations. the most favorite and powerful assessment is the parametric approaches using the AMMI model which generates Biplots to visualize stability and adaptability. However, this approach requires assumptions, namely normality, and homogeneity of variance. The Huehn method as a non-parametric approach based on genotype ranking does not depend on those assumptions. Evaluating the performance of the two approaches is very important to characterize their statistical properties. By a particular scheme of simulation, it can be evaluated the resistance of AMMI approaches to the presence of outlying observation. This study added 2%, 5%, and 10% outliers for genotypes across environments to rice and soybean data sets. Outliers were given by adding 3 times the standard deviation to the largest value in the randomly selected column/environment. It was found that the AMMI was sensitive to the presence of outliers even in the low number of outliers. The Huehn method is robust to the presence of outliers, but it tends to infer genotypes as stable by the conservative chi-square test. We propose to see the stability of each genotype relative to the others, using the rank of Z(1) and Z(2) indexes. Some genotypes which are relatively the most stable compared to others both based on Z(1) and Z(2) are similar.