Abstract:
|
Group testing, where subjects are tested in pools, is used widely to screen for infectious diseases. When testing involves a rare disease, pooled testing may increase the number of false negative test results due to pooled dilution effects. If testing results are used to estimate individual-level disease probabilities in a regression context, estimates of the regression parameters can be severely biased when the dilution effect is ignored. Most existing regression approaches assume that the assay sensitivity is a known constant, regardless of the pool size. In this project, we propose a new method for group testing data that adjusts for the potential dilution effect. We consider existing fixed effects and random effects regression models that have been proposed in the group testing literature. We specify a secondary parametric dilution model for the pooled-level sensitivity and estimate it together with the regression parameters from the primary regression model. Our approach provides reliable inference for regression parameters in the presence of a dilution effect. We illustrate our methods using hepatitis B data collected from a study involving Irish prisoners.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.