lofocv: Leave-one-function-out cross-validation

In refund: Regression with Functional Data

Leave-one-function-out cross-validation

Description

This internal function, called by fosr() when method="OLS", performs efficient leave-one-function-out cross-validation using Demmler-Reinsch orthogonalization to choose the smoothing parameter.

Usage

lofocv(Y, X, S1, argvals, lamvec = NULL, constr = NULL, maxlam = NULL)

Arguments

Y	matrix of responses, e.g. with columns corresponding to basis function coefficients.
X	model matrix.
S1	penalty matrix.
argvals	values where the functions are evaluated
lamvec	vector of candidate smoothing parameter values. If NULL, smoothing parameter is chosen by optimize.
constr	matrix of linear constraints.
maxlam	maximum smoothing parameter value to consider (when lamvec=NULL).

Value

if lamvec=NULL, a list (returned by optimize) with elements minimum and objective giving, respectively, the chosen smoothing parameter and the associated cross-validation score. Otherwise a 2-column table with the candidate smoothing parameters in the first column and the corresponding cross-validation scores in the second.

Author(s)

Philip Reiss [e-mail] and Lei Huang

See Also

fosr, pwcv

refund documentation built on Sept. 21, 2024, 1:07 a.m.