Estimation Procedures in Group Testing

Abstract

Estimation of proportion of defective items using group testing has been proven to be more precise (reduced MSE) and economical than individual testing. Seven point estimators have been explored in this project with their applicability and emphasis laid on the bias corrected and empirical Bayes estimators. Using an example, these two class of estimators have been shown to be superior to the maximum likelihood and classical Bayesian estimators. Four methods of obtaining group size have been examined with an in-depth description of group sizes obtained from optimizing the MSE and the method of adjusting the group size from one testing phase to another which are recommended. Finally, six interval estimates are explored with special consideration laid on the variance stabilized interval (VSI) a Wald interval generated with a stabilized variance (free of the proportion estimate). Using an example this Interval has been shown to overcome the problem of negative endpoints.