On the Effect of False Positives in Group and Individual Testing of Blood for the Presence of HIV

Abstract

In this thesis we develop a model which allows us to evaluate Modified Dorfman's and Modified Sterrett's group testing procedures in the presence of false positive classification error that can be used for screening blood for HIV. Performance measures derived are the expected number of tests per sample and the corresponding coefficient of variation. The procedures differ from the original ones by the fact that groups and samples are retested certain number of times before they are classified as positive. Modified Individual testing procedure and the procedure currently used by Canadian Red Cross are also evaluated and all four testing strategies are compared. Numerical analysis illustrates that group testing is more efficient than alternatives.