Dose-response estimation by smoothing and area under the curve

Abstract

Many current dose-response methods assume a parametric shape of some form and do not accommodate comparisons between methods. Motivated by a need to be able to compare dose-response curves without assuming a parametric shape to the dose-response curve, we applied our technique to model them by non-parametric smoothing methods and also estimate area under the curves. Given that we do not know the distribution of the area under the smoothed curve, we re-formulated the F-test to compare the areas between curves. We demonstrate this method on several actual datasets from immunology which have doseresponse data. This demonstration shows that this method can easily be employed to model dose-response curves and compare them between groups with multiple visits, which will provide a useful tool for this concept and hopefully in the future for dose-response relationships in gene mapping.