zheyuanli / gps.mgcv

General P-Splines for Package 'mgcv'

gps.mgcv (version 1.0)

General P-Splines for Package 'mgcv'

The package constructs and predicts the general P-splines of Li and Cao (2022) in mgcv by defining a new 'gps' smooth class (see ?gps.smooth). A general P-spline f(x) is specified as s(x, bs = 'gps', ...) in a formula and estimated using mgcv's model fitting functions, namely gam (generalized additive models, or GAMs), bam (GAMs for big data) and gamm (GAMs as mixed-effect models). General P-splines are state-of-the-art penalized B-splines. Unlike the standard P-splines of Eilers and Marx (1996) that only make sense for uniform B-splines on equidistant knots, they are properly defined for non-uniform B-splines on irregularly spaced knots, thanks to their powerful general difference penalty that accounts for uneven knot spacing. The package also contains functions for fitting and benchmarking different penalized B-splines (see ?Fit4BS and ?SimStudy) and a case study of smoothing Bone Mineral Content longitudinal data (see ?BMC).

Installation from CRAN

The package is not on CRAN yet, so you need to install it from this GitHub repository for now.

Installation from GitHub

The package requires R (>= 4.0.0), gps (>= 1.1) and mgcv (>= 1.8-40).

Package mgcv comes with an R distribution, but you may need to update it if your existing version is older than the requirement. You can always install its latest version from CRAN using install.packages("mgcv").

Install both gps and gps.mgcv from GitHub:

## you may need to first install package 'devtools' from CRAN
devtools::install_github("ZheyuanLi/gps")
devtools::install_github("ZheyuanLi/gps.mgcv")

Vignette

R code for Li and Cao (2022) is at: https://github.com/ZheyuanLi/gps-vignettes/blob/main/gps1.pdf