cv.spls: Compute and plot the cross-validated error for SPLS regression

Description

Draw the heatmap-type plot of the v-fold cross-validated mean square prediction error and return the optimal eta (thresholding parameter) and K (number of hidden components).

Usage

cv.spls( x, y, fold=10, K, eta, kappa=0.5,
        select="pls2", fit="simpls",
        scale=TRUE, center=TRUE, scale.y=FALSE )

Arguments

Matrix of predictors.

Vector or matrix of responses.

fold

Number of cross-validation folds. Default is 10-folds.

Number of hidden components.

eta

Thresholding parameter. eta should be between 0 and 1.

kappa

Parameter to control the effect of the concavity of the objective function and the closeness of the original and surrogate direction vectors. kappa is relevant only when the responses are multivariate. kappa

select

PLS algorithm for variable selection. Alternatives are "pls2" or "simpls". Default is "pls2".

fit

PLS algorithm for model fitting. Alternatives are "kernelpls", "widekernelpls", "simpls", or "oscorespls". Default is "simpls".

scale

Scale the predictors by dividing each predictor variable by its sample standard deviation?

center

Center the predictors?

scale.y

Scale the responses by dividing each response variable by its sample standard deviation?

Value

Invisibly returns a list with components:
mspematMatrix of the cross-validated mean squared prediction error. Rows correspond to eta and columns correspond to the number of components (K).
etaoptOptimal eta.
KoptOptimal K.

References

Chun, H. and Keles, S. (2007). "Sparse partial least squares for simultaneous dimension reduction and variable selection", (http://www.stat.wisc.edu/~keles/Papers/SPLS_Nov07.pdf).

Examples

Run this code

data(yeast)
set.seed(1)
# MSPE plot. eta is searched between 0.1 and 0.9 and
# the number of hidden components is searched between 1 and 10
cv <- cv.spls( yeast$x, yeast$y, K = c(1:10), eta = seq(0.1,0.9,0.1) )
# Optimal eta and K
cv$etaopt
cv$Kopt