Assist: A Suite of S Functions Implementing Spline-Smoothing Techniques

• Activity Number: 165
• Type: Invited
• Date/Time: Monday, August 4, 2008 : 2:00 PM to 3:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Yuedong Wang*+ and Chunlei Ke
• Companies: University of California, Santa Barbara and Amgen, Inc
• Keywords: smoothing spline ; R package ; reproducing kernel Hilbert spaces ; semi-parametric regression ; non-parametric nonlinear regression ; random effects

Abstract

We present a suite of user friendly R functions for fitting various smoothing spline models including (a) nonparametric regression models for independent and correlated Gaussian data, and for independent binomial, Poisson and Gamma data; (b) semiparametric linear mixed-effects models; (c) nonparametric nonlinear regression models; (d) semiparametric nonlinear regression models; and (e) semiparametric nonlinear mixed-effects models. The general form of smoothing splines based on reproducing kernel Hilbert spaces is used to model nonparametric functions. Thus, these R functions deal with many situations in a unified fashion. Well-known special cases are polynomial, periodic, spherical, thin-plate and L splines, GAM, SS ANOVA, projection pursuit, multiple index, varying coefficient, functional linear and self-modeling nonlinear regression models.