Evaluation of nearest neighbor methods in wheat genotype experiments

Spatial variability among experimental plots may be a relevant problem in field genotype experiments, especially when a large number of entries are involved. Whenever variability cannot be controlled by blocking, nearest neighbor methods can be helpful. The purpose of this study is to evaluate nearest neighbor methods and to present other control methods for intense soil heterogeneity. Three experimental trials with wheat varieties were analyzed by using the following spatial techniques: Papadakis method, modified Papadakis methods, moving average, first-order autoregressive and first differences. The results of the above analyses were compared with those of a traditional Randomized Complete Block (RCB) ANOVA. It can be concluded that, in the presence of soil heterogeneity, nearest neighbor analysis (NNA) improves the experimental accuracy, by significantly decreasing the residual (unexplained) variability. For these three experiments, the first-difference model was the most effective among all the tested NNA methods.