Spatial Corrections of Unreplicated Trials using a Two-dimensional Spline
- K. R. Robbins *,
- J. E. Backlund and
- K. D. Schnelle
For trials exhibiting within-field spatial variability, good predictions of genetic merit require models that accurately estimate and partition spatial effects from genetic sources of variation. Ideally such models should be robust, providing good solutions across a range of spatial patterns. To achieve this, a new method for spatial corrections is proposed, which utilizes two-dimensional spline (2DS) interpolation to estimate spatial effects. Simulation experiments were used to compare this method with two-dimensional separable autoregressive models (AR1) and mixed models containing marginal pass and range effects (PRM). Models were evaluated based on accuracy in partitioning genetic and spatial sources of variation as well as robustness across different spatial and genetic effect structures. Results showed the 2DS method to be the most robust of the three methods, yielding accurate spatial corrections across all simulation scenarios, whereas the AR1 model showed indications of over-fitting in the presence of genotype × environment interactions and the PRM model performed poorly in the presence of pass × range interactions. Applications using maize (Zea mays L.) hybrid yield and moisture data confirmed the effectiveness of the 2DS method at reducing the error variance by accounting for spatial trends. This new model provides breeders with a robust method to model within-field spatial variability across a range of experimental designs and spatial effect distributions.Please view the pdf by using the Full Text (PDF) link under 'View' to the left.
Copyright © 2012. . Copyright © by the Crop Science Society of America, Inc.