PriorNormPCA {mlica} | R Documentation |
This function performs a simple PCA analysis to aid in threshold setting and noise removal.
PriorNormPCA(X)
X |
Data Matrix (need not be normalised). Subsequent ICA seeks independent modes as independent distributions with values "down the rows". |
This function performs a simple PCA analysis and is used prior
to application of the main ICA algorithm. The objective of the prior
PCA is to help determine the dimensionality of a subspace on which
the further ICA converges. The convention used here is that the rows of X
label
the space over which independent components are sought. For a typical
microarray application in which ICA is being used as a generative
model for gene expression, rows should label genes and columns should
label samples. If, however, ICA is to be used as an unsupervised
projection pursuit algorithm, rows should label samples and columns
genes. For the latter application, the number of genes should be less than the
number of samples.
A list with following components:
X |
Normalised data matrix with the mean of each column set to zero. |
Dx |
Eigenvalues in a diagonal matrix. |
Ex |
Eigenvectors |
Andrew Teschendorff aet21@cam.ac.uk