A1328
Title: Nonparametric estimation of reference curves
Authors: Sandie Ferrigno - INRIA Nancy and University Nancy Lorraine (France) [presenting]
Abstract: In epidemiology, reference or standard curves are required to study fetal development in pregnancy. Values that lie outside the limits of these reference curves may indicate the presence of a disorder. Some classical empirical, parametric and semi-parametric methods, such as polynomial regression and LMS methods, are usually used to construct these curves. However, these classical methods build upon restrictive assumptions on estimated curves. The focus is on alternative nonparametric methods such as Nadaraya-Watson kernel estimation, local polynomial estimation, B-splines or cubic splines. The practical implementation of these methods to construct these curves requires working on smoothing parameters or choice of knots for the different types of nonparametric estimation. In particular, the optimal choice of these parameters is proposed. To fit these curves, we develop the R package quantCurves, an easy-to-use tool for practitioners, and a graphical interface to enable intuitive visualization of the results of the package.
