On the design of field experiments with correlated treatment effects
Large-scale field evaluation of genetic material forms an important part of the selection process in the early stages of plant breeding programs. These experiments are typically designed ignoring information on genetic relatedness, often available in the form of crossing history, or plant pedigree records. This paper considers the design of plant breeding experiments where the residuals may be correlated with an assumed autoregressive process, and there is a known genetic covariance structure among genotype effects. This structure is frequently more complex than simple nested family models, arising more generally from the pedigree, or possibly identity in state measures. It is widely accepted that the analysis of these data is improved using information on related individuals. The design of these experiments exploiting known genetic relatedness is considered using three case studies from industry that differ in selection goals, genetic complexity and scale.