Methods that reduce the cost and time involved in detecting defective or nonconformal members of a large population have been explored extensively in the quality control literature.
These methods have also found extensive application in insect-vector, rodent-bacterium and blood screening. Group-screening designs are plans that identity defect factors in a large population by initially pooling factors together and then classifying each pool as nonconformal (NC) or conformal (C). Individual testing is then carried only amongst individual factors in pool that are found to be nonconformal. A modifications of this strategy, suggested by Sterrett (1957), proposes a reversion to a group test, in a group declared defective, upon detection of the first nonconformal factor and then carrying out individuals testing only if the new group is nonconformal. This procedure is referred to as the Dorfman-Sterrett procedure in the literature.
The statistical properties of the restricted Dorfman-Sterrett procedure, where the number of reversion to a group test is predetermined, has found little discussion in the literature. This study uses a testing of hypothesis approach to compare the performance of the Dorfman-Sterrett procedure with the Dorfman procedure assuming that factors or groups can be misclassified. Under the testing of hypothesis approach, using a 2g fractional factorial design, cost functions which are linear functions of expected total number of incorrect decisions and the expected.