Analysis of residuals for stability assessment of rice genotypes

The analysis of residuals computes general adaptabilities (GAs) and specific adaptabilities (SAs) of genotypes similar to combining ability analysis. The residual analysis is simple and robust being based on the first-degree statistics, and is assumption-free. A unique statistic of ratio of variances (VR) of residuals for individual genotypes allows comparing their stability. The residuals analysis of genotype × environment (GE) interactions was performed for grain yield (t ha^?1) of 25 rice (Oryza sativa L.) genotypes, including two developed by marker-assisted selection for root traits quantitative trait loci (QTL), tested across 16 sites. Results were compared with the linear regression analysis for (i) full set of 16 sites showing adequacy of linear model and (ii) subset of nine-poor sites showing inadequacy of linear model. There was agreement between the two analyses for situation (i) but not for situation (ii) when the regression analysis fails because of nonlinearity but the residual analysis retains its validity.