Submitted to: Phytopathology
Publication Type: Abstract only
Publication Acceptance Date: 4/5/2006
Publication Date: 7/28/2006
Citation: Turechek, W. 2006. Statistical methodologies for early detection surveys in forest and ornamental landscapes. Phytopathology. 96:S144. Interpretive Summary:
Technical Abstract: In surveys designed for early detection of a pathogen it is typical for a large number of samples to be collected and brought to the laboratory for confirmatory, diagnostic testing, often with PCR or ELISA. When the incidence of the pathogen is expected to be low, group or batch testing is preferred to testing individuals for both practical and statistical reasons. Samples can be processed more rapidly and estimates of incidence are more precise with group testing when certain conditions are met. Under the usual binomial model group size, k, is often selected under the assumption that testing, and hence classification, occurs without error. In practice diagnostic tests are not perfect. Optimal k should be selected based on a test’s sensitivity (Se) and specificity (Sp) and the extent to which each is affected by grouping. A reduction in Se results in false negative classifications. This occurs when the target is diluted below the test’s detection threshold; it should be evident that the magnitude of this effect is highly dependent upon k. When the probability of target detection decreases as k increases, a dilution effect exists. Methods to account for a dilution effect under the binomial model will be presented. The impact of a dilution effect on the estimation of incidence and the selection of an appropriate group size when a dilution effect exists will be addressed using examples drawn from real-time PCR detection assays for <i>Xanthomonas fragariae</i>, casual agent of angular leaf spot of strawberry, in strawberry nursery stock.