GroupTesting: an R package for group testing estimation

Abstract

Group (pooled) testing has long been used for the monitoring and detection of infectious diseases. In group testing, pools comprised of individual biospecimens are amalgamated and tested initially, and then the individuals belonging to positive pools are retested to complete case identification. This procedure and its variants can offer substantial savings with regard to time and testing cost. Unfortunately, these savings come at the expense of a complex data structure; e.g., ambiguities due to imperfect testing and complex dependence caused by individuals potentially being tested in multiple pools. To account for these complex features, several advanced statistical methods have been proposed. Regretfully, these methods are non-trivial to implement, especially for non-statistician. Recognizing this as an important issue in surveillance programs, we have developed a user-friendly R package called groupTesting that can be used to analyze group testing data. In particular, our package consists of R functions which possess a great deal of versatility and generality, and can be used in estimating both proportions and binary regression functions. The computing efficiency of our package is greatly enhanced through the strategic implementation of compiled Fortran subroutines. The primary features of the R package are illustrated using HIV and chlamydia data.