Purpose: The goal of this study was to develop a simple tool for evaluating the performance characteristics of three commonly used preharvest sampling plans: random, stratified-random, and z-pattern sampling.
Methods: Mathematical derivations and computer simulations by Matlab were performed to compare the relative effectiveness of random, stratified-random, and z-pattern sampling. This initial evaluation included consideration of both the number of contamination sites in the field and the number of samples analyzed.
Results: The detection probability with increasing number of contamination sites and increasing number of sample. When a large number of simulations were performed, the results agree well with the theoretical derivations, i.e., there was no difference on the mean detection probability for the three sampling plans. The detection probability decreased rapidly as a function of total number of subplots tested, in roughly a linear manner on log-log plots. However, the inherent variability of the z-pattern sampling plan was substantially greater than the other sampling plans. This difference is most dramatic when the number of contamination sites is small, with the z-pattern sampling plan having a high frequency of no-detection responses. Thus, the z-pattern sampling plan inherent variability has an increased probability of type 1 and type 2 errors compared to either random or stratified-random sampling plans.
Significance: This study provides a simple mathematical approach for evaluating the effectiveness three commonly used preharvest sampling plans, and suggest that with low level contamination, random or stratified-random sampling plans would likely be more effective.