The influence of error variance variation on analysis of genotype stability in multi-environment trials

Multi-environment trials are often analyzed to assess the yield stability of genotypes. Different approaches to stability analysis can be cast into a unifying mixed modeling framework. The choice of the class of candidate mixed models and the estimates of the model parameter have direct implications for the stability measure. The heterogeneity of residual error variances across environments generally exists in multi-environment trials. The objectives of this study were to investigate the impact of the analytical procedure with different considerations about error variances when assessing yield stability of genotypes. A series of 16 multi-environment trials from a corn-breeding program in the north of China were simultaneously analyzed from 2005 to 2008 using a randomized complete block design at each environment; the analysis used five most common stability models with homogeneous residual error variances, as well as their heterogeneous residual error variance versions to take into account that different environments may have different levels of precision. The results showed that whether the error variance differences across environments were accounted for in the analysis procedure did not affect the choice of appropriate models for stability analysis, but considerably influenced the estimates of model parameters (percentage difference of the parameter estimates between models with heterogeneous and homogeneous residual error variances varied from −122.4% to 65.7% depended on genotypes and trials), and hence influenced the stability ranking of some genotypes. The models with heterogeneous residual error variances fitted the trial data better and gave (with 2.1-8.4% reduction) smaller standard errors of model parameter estimates than their homogeneous residual error variance versions, which suggests that the model with heterogeneous residual error variances constitutes a good alternative analysis for genotype stability in multi-environment trials.