MOVING MEAN AND LEAST-SQUARES SMOOTHING FOR ANALYSIS OF GRAIN-YIELD DATA

Soil heterogeneity often decreases precision in large yield trials. Estimation of, and adjustment for, fertility trends within a trial may increase precision. Two methods of estimating fertility trends were evaluated in 12 experiments with hexaploid what (Triticum aestivum L.) and 11 experiments with tetraploid wheat (T. turgidum L. var durum). Each experiment consisted of a trial with two replications of 114 to 282 entries repeated at each of two local:ions in south-western Saskatchewan, Canada. With moving mean, the fertility trend for each plot was calculated as the average of the unadjusted yields of six neighbor plots. With least squares smoothing, the fertility trend was calculated as a weighted average of plot yields adjusted for differences among treatments. Weights were inversely proportional to the distance from the adjusted plot and depended upon the observed fertility trend. Compared with unadjusted yields, adjustments by moving means increased the correlation between the two trials within an experiment from an average of 0.41 to 0.48, whereas least squares smoothing increased the average correlation from 0.41 to 0.49. These increases indicate an average increase of within-trial precision of 20% for moving mean analysis and 24% for least squares smoothing. Either method is useful for removing fertility trends in large yield trials.