The precision of estimates of genotype means in field trials can be increased by using an appropriate experimental design and spatial modelling techniques. Both design-based and spatial analyses are usually based on the assumption of homogeneous variance. But in reality this assumption may not generally hold true. If this is ignored, imprecise fixed effect estimates can result. The aim of this study is to provide analyses accounting for possible variance heterogeneity along with the spatial trend if any. The methods are explored using three maize trials from Ethiopia. We consider the Box-Cox transformation to stabilize variance, as well as the power of the mean and exponential variance models to allow for heterogeneity. The best model for each datasets are selected using AIC.
The Box-Cox transformation was found to be successful in stabilizing the variance but inferring of results on the original scale was difficult. The heterogeneous variance models avoid this problem. In two out of three datasets, the spatial model along with the variance models improved the fit, while in the third data set the design based analysis along with variance model had the best fit criteria.
|