Analysis of Virus Local Lesion Experiments

Calculations based on variation of lesion counts about means within the range 7 to 260 per half‐leaf show that the standard deviation may be rendered independent of the mean by a transformation log (x+12) for half‐leaf experiments in bean, tobacco and Nicotiana glutinosa and log (x+21) for whole‐leaf experiments in bean. Half‐leaf experiment designs are superior to those using whole‐leaves. Balanced Incomplete Block designs are more efficient than Randomized Blocks or comparisons using a common standard. The use of covariance to remove the effect of leaf susceptibility gives increased accuracy over the treatment‐standard comparison if used with designs allowing removal of both plant variation and that due to leaf age.