ASSOCIATION BETWEEN PARAMETRIC AND NONPARAMETRIC MEAURES OF PHENOTYPIC STABILITY IN RICE GENOTYPES (Oryza sativa L.)

Abstract

Evaluation of performance stability and high yield is essential for yield trials in different environments. The mostly used, classical parametric approaches for an analysis of genotype x environment interaction are based on several assumptions: normality of the distribution, homogeneity of variances, additively. If some of mentioned assumptions are not fulfilled, the validity of these methods may be questionable. By use of nonparametric methods, which are simple and easy for analysis, all of the mentioned assumptions are avoided. In this paper we used five of parametric and 11 of nonparametric techniques for analysis of genotype x environment interaction for grain yield of 7 rice (Oriza sativa L.) genotypes through three locations in two years (2005, and 2006). The objectives of this study were to study the interrelationship among various parametric and nonparametric phenotypic stability statistics, and to evaluate the similarity between these methods, and to determine the most suitable methods for assessing the rice genotypes yield stability. Values of the stability measures shown that genotypes with the highest grain yield in the majority of cases were not the most stable. The results of Spearman’s rank correlation indicates that the nonparametric measure Si(1), Si(2), Si(3), Si(6), NP(1), NP(2), NP(3), NP(4) and parametric measures bi, S2d, S2i, and Wi were positively related with each others and negatively correlated with mean yield and only the rank-sum and modified rank-sum showed a positive correlation with mean yield. The Principle Component Analysis (PCA) showed four distinct groups: group1 consist of Si(1), Si(2), Si(3) ,Si(6), NP(1), NP(2), NP(3),NP(4),bi,,S2d,S2i, and Wi ; group 2 consist of RS, RS1, and RS2; group 3 consist of mean yield (Y); and group 4 consist of CV. In conclusion, the modified rank-sum methods (Rs1, Rs2) which use the nonparametric measures Si(1) and Si(2) with the rank mean yield of genotypes, seems to be useful under conditions where the basic assumptions of parametric stability are not met, and for simultaneously selection for high yield and stability.