Estimation of genotypic values without replication in field trials

Some properties of a certain class of local adjustment methods, aiming at correction for spatial variation in growing conditions in trial fields, have been studied. In particular, the precision of the moving average estimator (MAE) and the best linear unbiased estimator (BLUE) for genotypic values in mass selection were compared. Under quite general assumptions about the nature of spatial heterogeneity, it appeared that the BLUE will in most cases be hardly better than MAE. With respect to the neighbourhood size to be chosen for calculation of the local moving mean, a circular area with a radius of slightly over twice the interplant distance was found to be satisfactory. When the covariance structure of the trial field is unknown, as will generally be the case, this figure seems to be a safe guideline for the size of a moving grid.