new.penalized.pls {ppls}R Documentation

Prediction for Penalized Partial Least Squares

Description

Given an object returned by the function penalized.pls, this function predicts the response for new data, with different numbers of components (as specified when running penalized.pls).

Usage

new.penalized.pls(ppls, Xtest, ytest = NULL)

Arguments

ppls An object returned from penalized.pls
Xtest A matrix of new input data with the same number of columns as the matrix X given as an argument of the function penalized.pls to construct ppls.
ytest The numeric vector of new response data, optional. Its length must equal the number of rows of Xtest.

Details

The function penalized.pls simply returns the intercepts and regression coefficients for all penalized PLS components up to ncomp which is specified as an argument. The function new.penalized.pls then computes the estimated response using these regression vectors. If ytest is given, the mean squared errors for all number of components are computed as well.

Value

ypred The matrix of predicted responses. The number of rows equals the number of rows of Xtest, whereas the number of columns equals the ncomp argument given to the function penalized.pls while constructing the object ppls.
mse vector of mean squared errors, if ytest (giving the true response) is provided.

Author(s)

Nicole Kraemer

References

N. Kraemer, A.-L. Boulesteix, G. Tutz (2007) "Penalized Partial Least Squares with Applications to B-Splines Transformations and Functional Data", preprint

available at http://ml.cs.tu-berlin.de/~nkraemer/publications.html

See Also

penalized.pls, penalized.pls.cv, ppls.splines.cv

Examples

# see also the example for penalised.pls 
X<-matrix(rnorm(50*200),ncol=50)
y<-rnorm(200)
Xtrain<-X[1:100,]
Xtest<-X[101:200,]
ytrain<-y[1:100]
ytest<-X[101:200]
pen.pls<-penalized.pls(Xtrain,ytrain,ncomp=10)
test.error<-new.penalized.pls(pen.pls,Xtest,ytest)$mse

[Package ppls version 1.0 Index]