gsir: gsir

In nsdr: Nonlinear Sufficient Dimension Reduction

Description

gsir

Usage

gsir(x,x.new,y,ytype,ex,ey,complex.x,complex.y,r)

Arguments

x	input predictor matrix from training set
x.new	input predictor matrix from testing set
y	response variables
ytype	type of response variables
ex	tuning parameter for the Tychonoff regularized inverse for GX
ey	tuning parameter for the Tychonoff regularized inverse for GY
complex.x	tuning parameter for the Gaussian kernel in X
complex.y	tuning parameter for the Gaussian kernel in Y
r	number of dimension

Value

suff.pred: sufficient predictors from GSIR

obj.mat: objective matrix of GSIR

eig.val: the first r eigenvalues from the eigendecomposition of the objective matrix

eig.vec: the first r eigenvectors from the eigendecomposition of the objective matrix

References

Li, B. (2018). Sufficient dimension reduction: Methods and applications with R. CRC Press.

Examples

n = 50; p = 5; sigma = 1;
x = matrix(rnorm(n*p),n,p) ; err = rnorm(n)
y = sin(0.5+(x[,1]+1)^2) + sigma*err; ex=0.01 ; ey=0.01
gsir_res <- gsir(x,x,y,"scalar",ex,ey,1,1,1)

nsdr documentation built on June 3, 2021, 9:06 a.m.