new_cpc: FG Algorithm

In bbolker/cpcbp: common principal components/back-projection analysis

Description

Applies the FG Algorithm (Flury and Gautschi, 1986) to a set of matrices to be simultaeneously orthogonally diagonalized by a common matrix B. The matrix B contains the common principal components (CPC).

Usage

  new_cpc(matList, n = rep(1, length(matList)),
    verbose = FALSE, eps.F = 1e-06, eps.G = 1e-06,
    Bstart = diag(p))

Arguments

matList	list of covariance matrices
n	numeric vector of sample sizes
eps.F	accuracy constant for determining when to stop the F-level
eps.G	accuracy constant for determining when to stop the G-level
Bstart	initial B approximation (by default diag(p))

References

Flury B.N.; Gautschi, W. An algorithm for simultaneous orthogonal transformation of several positive definite symmetric matrices to nearly diagonal form. SIAM J. Sci. Stat. Comput. 1986, 7, 169-84

Examples

source("test_examples.R")
new_cpc(test_iris)

bbolker/cpcbp documentation built on May 11, 2019, 9:28 p.m.