Developing an R-library for spatial analysis of experiments

Description of the topic

Incomplete block designs and alpha-lattice field designs have helped increase the precision of estimating genetic values of breeding individuals. However, blocks or incomplete blocks do not capture all the plot-to-plot variability in a field experiment; therefore, appropriate modeling of the error term in field trials is indeed necessary. Analytical models that model the spatial variability considering the grid location of plots in the field have improved the precision of estimating the treatment mean, which produces greater genetic gains in plant breeding than when this spatial adjustment is not performed. A limited number of expensive commercial software (SAS, ASReml, Genstat) and specifically one of the most used spatial models, the separable autoregressive in the direction of the rows and the columns, can model this spatial variability. However, no free software is available for this analysis. Some efforts have been made in R but the results are not reliable. Free software for spatial analysis of experiments will provide breeders with more precise genetic values at minimum cost.

Work expectations

The main objective of this proposal is to develop a free R-library for the spatial analysis of experiments. A review of existing R libraries for the analysis of experiments should be performed to identify the most suitable library, which will then be modified to include the modeling of the error by a spatial model. If no suitable library is found, a new R package including spatial analysis of field trials must be developed. Other activities included in the project are to produce technical documents for scientists and plan training activities to disseminate the new library.

Required skills

Linear mixed models and computing programming in R.