EFFICIENT CONTROL OF SPATIAL VARIATION IN YIELD TRIALS USING NEIGHBORPLOT RESIDUALS

The effect of spatial variation on experimental error variance and the significance of differences between genetic entries was evaluated in five performance trials. A significant portion ofspatial variation cou ld be detected in all the experiments investigated and various neighbo ur plot residuals were applied to adjust for local field trends. Neigh bour plot adjustment was clearly more efficient than the randomized co mplete block design in reducing error variance and in detecting signif icant differences between entries. It was also more efficient than lat tice designs in trial fields exhibiting short distance trends, which c ould not be covered efficiently by incomplete blocks. In most experime nts with long and narrow plots, longitudinal adjustment using two neig hbours at each side of a test plot was superior to adjustment by only one 'nearest' neighbour.