The coefficient of variation (CV) was created to measure population variability. However, its most common use is to measure validity of field experiments. The CV can be used to measure variability in genetic populations, to determine the best plot size in uniformity trials, to measure stability of phenotypes, or measure variation in other individual or population attributes. The CV is based on the assumption that the mean and the error variance change together such that regressing the natural log of error variance on the natural log of the mean produces a β = 2.0. Twenty-two sets of crop performance data revealed no relationship approaching β = 2.0. Thus, there is no basis for using the CV for crop performance trials. If concern exists about heterogeneity of error variances, the data could be transformed on the basis of the relationship between the error and mean for that crop. Locations with higher yield have a larger impact on overall means than locations with lower yield, but one could minimize that impact by calculating relative yield. The CV should no longer be used to indicate validity in most field trials, particularly crop performance trials.
