pcaMarchenkoPastur: Marchenko-Pastur Significant PCs

In farrellja/URD: URD

Description

The Marchenko Pastur Law (MP) predicts the theoretical upper and lower bounds on the null distribution of eigenvalues for an MxN random matrix. We take significant principal components (PCs) as those with eigenvalues greater than the maximum eigenvalue predicted for random data. This function assumes that the data has mean 0 and variance 1 (i.e. that the data has been centered and scaled). This called automatically by calcPCA and the results are stored in slot pca.sig.

Usage

pcaMarchenkoPastur(M, N, pca.sdev, factor = 1, do.print = T)

Arguments

M	(Numeric) Number of rows in input data
N	(Numeric) Number of columns in input data
pca.sdev	(Numeric vector) Standard deviations for each principal component
factor	(Numeric) Factor to multiply eigenvalue null upper bound before determining significance.
do.print	(Logical) Whether to report the results

Value

Logical vector of whether each PC is significant.

farrellja/URD documentation built on June 17, 2020, 4:48 a.m.