Abstract:

The influence of disease incidence, clumping of infected plants, and size of the sampling unit on the sensitivity of the ordinary runs test was simulated in order to identify the optimum sampling profile for investigating the spatial pattern of plant diseases. A simulation programme was written to generate plant populations with random and clustered spatial patterns displayed in quadrats of 50 by 200 plants. 600 and 1200 independent populations were generated for random and clustered patterns, respectively. 12 levels of disease incidence were simulated within the range 0.01-0.95. For every population, the simulation performed a sampling procedure with 9 sizes of the sampling unit (20, 30, 40, 50, 60, 80, 100, 120 and 150 plants). In each case the simulation was based on 1000 samples of continuous series of plants. In order to evaluate the sensitivity of ordinary runs test to the degree of aggregation, plant populations were simulated with two additional values of clumping power, for two levels of disease incidence. When a random pattern was simulated, the probability of rejecting the null hypothesis was almost unaffected by the size of the sampling unit and slightly decreased with disease incidence. When clustered patterns were generated, the probability of error clearly decreased both with disease incidence and size of the sampling unit. The probability of error was also affected by the degree of aggregation. As expected, the higher the clumping power the higher the probability of rejecting the null hypothesis. The implications of the sensitivity of the runs test on the design of sampling schemes are discussed.