AN EXAMINATION OF FIVE SAMPLING METHODS UNDER RANDOM AND CLUSTERED DISEASE DISTRIBUTIONS USING SIMULATION

Two fields with random disease distribution and eight fields with clustered disease distribution were simulated at each of five disease incidence levels (10, 20, 30, 40 and 50% infected plants). Five sampling methods differing with respect to field area covered (whole field vs. demarcated area of 1/8 of the field), and shape of path (X, W, and diagonal) were applied to these fields. Each path was considered to have four arms and two sampling sizes were used, 3 sites and 10 sites per arm. The variation within field showed that arm effects were significant for clustered disease distribution with 10-site sampling. Increasing sample size within arms decreased the coefficient of variation percentage and increased the precision of estimates for random disease distribution, but no similar improvement was observed for clustered disease distribution. The precision was of the same magnitude for all methods under the random disease conditions; however, it differed considerably for clustered distributions: in general the magnitude of variability for an X- or W-shaped path covering the entire field was about one-half of that for a diagonal path and about one-quarter of that for the methods covering only a portion of the field.