gplsfitGCV: Generalized Penalized Least Square Fit under GCV

In funstatpackages/GgAM: Generalized geoAdditive Models

Description

This is an internal function of package ggam.

Usage

gplsfitGCV(Y, B, Q2, P, UB = NULL, lambda, family, offset, theta = 0,
  fx, control, start = NULL, etastart = NULL, mustart = NULL,
  X = NULL, ind_c = 1:ncol(X), fixedSteps = (control$maxstep + 1),
  ...)

Arguments

Y	Response variable.
B	The bernstein basis matrix.
Q2	The Q2 matrix from QR decomposition of the transpose of the constraint matrix.
P	The penalty matrix.
UB	The univariate basis function matrix constructed.
lambda	The smoothing penalty parameter.
family	The family object, specifying the distribution and link to use.
offset	Can be used to supply a model offset for use in fitting. Note that this offset will always be completely ignored when predicting.
theta	The given theta values in negative binomial family.
fx	indicates whether the term is a fixed d.f. regression
control	A list of fit control parameters to replace defaults returned by plbpsm.control. Any control parameters not supplied stay at their default values. spline (TRUE) or a penalized regression spline (FALSE).
start	Initial values for model coefficients
etastart	Initial values for linear predictor.
mustart	Initial values for the expected response.
X	The parametric model matrix. set to 'NULL' if it is not provided.
ind_c	The vector of index to indicate the parametric part.
fixedSteps	How many steps to take: useful when only using this routine to get rough starting values for other methods.
...	other arguments.

Details

See section 4 'Implementation' in Shan et al. (2018).

Value

A list of fit information.

Examples

library(BPST)
data("eg_poi")
eg1_V1 <- eg_poi[['V1']]
eg1_T1 <- eg_poi[['T1']]
sam <- eg_poi[['sam_poi']]
d <- 2
r <- 1
B0 <- basis(eg1_V1,eg1_T1, d, r, sam[,c('loc1','loc2')])
B <- B0$Bi
ind <- B0$Ind.inside
Q2 <- B0$Q2
K <- B0$K
Z <- sam[ind,c(5:12)]
Y <- sam[ind,'y']
lambda_start <- 0.01
lambda_end <- 32
nlambda <- 10
lambda <- exp(seq(log(lambda_start),log(lambda_end),length.out=nlambda))
X <- sam[,1:15]
P <- t(Q2)%*%K%*%Q2
gplsfitGCV(Y,as.matrix(B),Q2,P,UB=NULL,lambda=lambda,family=poisson(),offset=0,fx=FALSE,
control = plbpsm.control(),X=as.matrix(X))

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