Submitted to: Crop Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/22/2008
Publication Date: 8/11/2009
Citation: So, Y., Edwards, J.W. 2009. A Comparison of Mixed-Model Analyses of the Iowa Crop Performance Test for Corn. Crop Science. 49:1593-1601.
Interpretive Summary: Identifying plant varieties (commercial and experimental) that provide superior performance for growers across a wide range of growing conditions is a very expensive task. Varieties must be evaluated in many environments and performance data summarized in order to identify not only the best varieties, but also those with the most stable performance. Several statistical methods of analyzing crop performance data were compared in order to determine the best comparisons among varieties can be obtained by assuming each test-site is equally informative. We found strong statistical evidence that comparisons among varieties could be improved by unequal weighting of information from different test sites. The results of this study will guide future studies to identify the best statistical models for comparing varieties from multiple test sites.
Technical Abstract: Multi environment trials (MET) are one of important steps in plant breeding toward cultivar selection and recommendation. The data generated by MET have inherent features from statistics perspectives that the data are often incomplete by selection and the typical statistical assumptions for MET data analyses are not reasonable. These issues can be addressed through likelihood-based mixed model approach. In this study, we compared mixed effect models that incorporated different types of heterogeneity of genotype variance, correlation and error variance in terms of goodness-of-fit using information criteria. Four different genotype variance covariance structures namely compound symmetry, heterogeneous compound symmetry, heterogeneous correlation, unstructured covariance structure were considered under homogeneous and heterogeneous error variance structures. The results showed that heterogeneity of error variance is the most important factor in improving model fit in our example data sets from Iowa Crop Performance Test for corn. Compound symmetry and unstructured covariance structures under heterogeneous error variance were the best fit model judged by AIC and BIC, respectively. The two common information criteria agreed on a same model a little more 50% of the time but in the other cases, we observed a tendency that AIC picked more complex modes while BIC chose simpler models. This study demonstrates the need of incorporating some types of variance heterogeneity in such MET data analyses.