ISPAM September 6 2014 Meeting

(However, it might be better to neither test nor harvest a suspect area) The idea is to make a judgment based on the harvested, or to be harvested produce. The sample should represent the lot (acre, 4 acre plot, etc.). Sampling that is concentrated in suspect areas would be a biased sample that would fail at a higher rate than a random sample of the entire lot. 5.2. Size and quantity of samples collected For the binomial distribution, the number of samples is provided by the sampling plan AQL, RQL, etc. discussed in sections 4.1 and 4.2. The unit that is considered in the AQL is the size of the sample. Thus, a sampling plan would provide, for example, a 90% chance of rejecting a lot in which 1% of 30 grams samples are contaminated. A natural size might be a serving size. (You folks who know the test method would have to tell me if this is unreasonable). From a Poisson distribution standpoint, the outcome is positive/negative for the area or weight of the sample. For a given total weight of sample, the larger the number of test portions, the more the sample will be representative (a better random sample) of the lot. Obviously, 60-5 gram leaves from across the acre will be more representative than a single 300 gram head. 5.3. Which part(s) of the produce to collect The answer to this is comparable to that in 5.1. A random sample of leaves in which each leaf that is being or would be harvested (there is no reason to sample the outer leaves if they are not harvested), has the same probability of being drawn, provides a reliable estimate of the contamination of the harvest. If the contamination is not random on the plant and the plant is sampled predominantly from areas that are more likely to be contaminated, the sample will be biased and fail at a higher rate than a random sample of all leaves. 5.4. Individual vs. composite testing Composite testing is fine provided the probability of detecting the 1 positive individual sample in the composite sample is the same as the probability of detecting the positive individual sample when tested individually. 5.5. Reaction to test results 5.5.1. Statistical meaning of a detected positive A culture positive provides evidence that the lot has a higher level of contamination than a lot that is at the AQL. If the lot were at the AQL, the probability of a culture positive is the value of the Type 1 error value (Often chosen at 5%). 5.5.2. Statistical meaning of a non-detect