|
Abstract:
|
Group testing, as a cost-effective strategy, has been commonly used to perform large-scale screening for rare infectious. Recently, the use of multiplex assays has transformed its goal from detecting a single disease to diagnosing multiple infections simultaneously. Previous work on multiple-infection group testing data mainly focused on the estimation of disease prevalence where a multinomial model was assumed and individual level information was not considered. To incorporate such information, difficulties arise under multinomial model, where the marginal inference for each infection cannot be derived. To remedy this drawback, we propose a parametric framework that originates from the context of multivariate logistic model which allows us to jointly model multiple-infection group testing data and also to draw interpretable regression results including estimation, hypothesis testing, and variable selection for each infection separately. We illustrate the finite sample performance of our method through simulation and by applying it to chlamydia and gonorrhea screening data collected from the Infertility Prevention Project.
|
