grplsfit: Variable selection with Bivariate penalized Spline (GCV/CV)...

In funstatpackages/GgAM: Generalized geoAdditive Models

Description

This is an internal function of package ggam. Bivariate penalized least squares problem is solved with penalty parameter chosen by GCV or CV. Variable selection by using adaptive LASSO or group SCAD is applied in parametric coefficients.

Usage

grplsfit(G, criterion, method, family, ind_c, VS, MI, ...)

Arguments

G	An object of the type returned by plbpsm when fit=FALSE.
criterion	The criterion to choose the penalty parameter lambda. "GCV" to use generalized cross validation method and "CV" for cross validation
method	'ALASSO' or 'SCAD' to penalize the coefficients for parametric part.
family	The family object, specifying the distribution and link to use.
ind_c	The given index of covariates that are selected.
VS	'TRUE' for using ALASSO/SCAD to select linear variables.
MI	Whether model identification is conducted or not.
...	other arguments.

Details

This is an internal function of package ggam. We propose a coordinate descent based algorithm to perform the variable selection efficiently. The smoothing penalty parameter could be chosen by GCV or CV using the routines: plsfitGCV and plsfitCV. In this function, the user can also choose whether to do variable selection or not.

Value

A list of fit information.

funstatpackages/GgAM documentation built on Nov. 4, 2019, 12:59 p.m.