RRS.cv: Reduced-rank ridge regression with rank and tuning parameter...

In xliu-stat/NRRR: Multivariate Functional Regression via Nested Reduced-Rank Regularization

Description

This function performs reduced-rank ridge regression with the rank and the tuning parameter selected by cross validation.

Usage

RRS.cv(Y, X, nfold = 10, rankmax = min(dim(Y), dim(X)), nlam = 100,
        lambda = seq(0, 100, length = nlam), norder = NULL,
        nest.tune = FALSE, fold.drop = 0)

Arguments

Y	response matrix.
X	design matrix.
nfold	the number of folds used in cross validation. Default is 10.
rankmax	the maximum rank allowed.
nlam	the number of tuning parameter candidates. Default is 100.
lambda	the tuning sequence of length nlam.
norder	a vector of length n that assigns samples to multiple folds for cross validation. Default is NULL and then norder will be generated randomly.
nest.tune	a logical value to specify whether to tune the rank and lambda in a nested way. Default is FALSE.
fold.drop	the number of folds to drop. Default is 0.

Value

The function returns a list:

cr_path	a matrix displays the path of model selection.
C	the estimated low-rank coefficient matrix.
rank	the selected rank.
lam	the selected tuning parameter for ridge penalty.

References

Mukherjee, A., & Zhu, J. (2011). Reduced Rank Ridge Regression and Its Kernel Extensions. Statistical analysis and data mining, 4(6), 612–622.

xliu-stat/NRRR documentation built on Jan. 9, 2021, 3:23 p.m.