groupPLS: Two-step group penalized partial least squares regression

In jialiwang1211/groupPLS: Two-step group penalized partial least squares regression

Description

Performs group penalized partial least squares regression.

Usage

groupPLS(formula, scale.X = FALSE, scale.Y = TRUE, k,
  group.index = NA, group.pen = TRUE, lambda.min = 0.01,
  lambda.max = 0.1, ncomp = 20, seed = 1)

Arguments

formula	model formula, y~x, where y are the response variables and x are the predictors.
scale.X	scale x by the sample standard deviation of each variable. TRUE or FALSE
scale.Y	scale y by the sample standard deviation of each variable. TRUE or FALSE
k	number of groups. If group.index is not provided, then perform the hierarchical clustering based on a similarity matrix to define k clusters.
group.index	pre-defined group index for each predictor.
group.pen	group penalization on latent components. TRUE or FALSE.
lambda.min	lower bound of group penalty paramter.
lambda.max	upper bound of group penalty paramter.
ncomp	number of latent components in partial least squares.
seed	an integer.

Details

More details.

Value

Return groupPLS object.

ncomp.group	indices of latent components from PLS in each group.
latent.group	combined latent components from PLS in each group.
coef	regression coefficients of latent components on responses.
scale.X	if scaling of X is requested.
scale.Y	if scaling of Y is requested.
Xmean	mean of each variable in X.
Xsd	standard deviation of each variable in X.
Ymean	mean of each variable in Y.
Ysd	standard deviation of each variable in Y.
k	number of groups.
group.index	group index for each predictor.
group.lambda	group penalty paramter selected by cross validation.
call	function call.

Author(s)

Jiali Wang ([e-mail]); Le Chang ([e-mail])

Examples

data<-read.csv("Spectra_eps.csv")
data<-data[complete.cases(data),]
rs<-groupPLS(data[,c(2,4:6)] ~ data[,7:dim(data)[2]], k=3)

jialiwang1211/groupPLS documentation built on Feb. 17, 2020, 12:35 a.m.