Jackknife estimation for PPLS-coefficients

Share:

Description

This function computes the mean and the covariance of the regression coefficients of Penalized Partial Least Squares.

Usage

1
jack.ppls(ppls.object,ncomp,index.lambda)

Arguments

ppls.object

an object returned by penalized.pls.cv

ncomp

integer. The number of components that are used. The default value is the cross-validation optimal number of components.

index.lambda

integer. The index of the penalization intensity, given by the vector lambda that was provided to penalized.pls.cv. The default value is the cross-validation optimal index.

Details

The function needs an object returned by penalized.pls.cv. It estimates the mean and the covariance of the regression coefficient (with ncomp components and penalization intensity indexed by index.lambda). This is done via a jackknife estimate over the k cross-validation splits. We remark that this estimation step is not discussed in Kraemer, Boulesteix and Tutz (2008).

Value

The function returns an object of class ”ppls”.

mean.ppls

The mean of the regression coefficients over all cross-validation splits. This is a vector of length ncol(X). Note that in general, this differs from the regression coefficients computed on the whole data set, but if the number of observations is fairly large, the difference should be small.

vcov.ppls

The covariance matrix of the regression coefficients. This is a symmetric matrix of size ncol(X) x ncol(X).

index.lambda

Index for the value of lambda that determines the regression coefficient.

ncomp

Number of components that determines the regression coefficients.

k

The number of cross-validation splits. These can be used to construct a t-test for the coefficients.

Author(s)

Nicole Kraemer

References

N. Kraemer, A.-L. Boulsteix, and G. Tutz (2008). Penalized Partial Least Squares with Applications to B-Spline Transformations and Functional Data. Chemometrics and Intelligent Laboratory Systems 94, 60 - 69. http://dx.doi.org/10.1016/j.chemolab.2008.06.009

See Also

penalized.pls.cv,ttest.ppls

Examples

1
2
3
4
5
6
data(cookie) # load data
X<-as.matrix(cookie[,1:700]) # extract NIR spectra
y<-as.vector(cookie[,701]) # extract one constituent

pls.object<-penalized.pls.cv(X,y,ncomp=10,kernel=TRUE) # PLS without penalization
my.jack<-jack.ppls(pls.object)