fscca: Fast sparse canonical correlation

In pimentel/fscca: Fast SCCA

Description

This is a fast implementation of sparse canonical correlation analysis.

Usage

fscca(X, Y, penalty_x, penalty_y, lam_x, lam_y, k_folds = 5L,
  n_components = 1L, center = TRUE, scale = FALSE)

Arguments

X	a matrix of dimension n x p
Y	a matrix of dimension n x q
penalty_x	a string indicating the penalty function to use on matrix X. Currently only "lasso" is implemented.
penalty_y	a string indicating the penalty function to use on matrix Y. Currently only "lasso" is implemented.
lam_x	a numeric vector of tuning parameters on X
lam_y	a numeric vector of tuning parameters on Y
k_folds	an integer denoting the number of folds to use in cros-validation
n_components	the number of components to compute
center	center the columns to mean zero?
scale	scale the columns to standard deviation one?

Value

A list with matrices:

A

A matrix of dimension p x n_components of the canonical vector

B

A matrix of dimension q x n_components of the canonical vector

U

A matrix of dimension n x n_components of the X * a

V

A matrix of dimension n x n_components of the Y * b

lambda

A matrix of dimension n_components x n of the optimal tuning parameters

covar

The average cross-validated covariance

pimentel/fscca documentation built on May 25, 2019, 7:12 a.m.