Signal Matrix Extraction for DIVAS

Extracts the signal matrix for each data block, estimates signal ranks, and adjusts perturbation angles for each data matrix.

Usage

DJIVESignalExtractJP(
  datablock,
  nsim = 400,
  iplot = FALSE,
  colCent = F,
  rowCent = F,
  cull = 0.382,
  noisepercentile = rep(0.5, length(datablock)),
  noiselvls = NULL
)

Arguments

datablock

A list of feature by sample (eg gene by cell) data matrices, each matrix corresponding to a data block. Matrices have to have same number of columns (mathced samples).

nsim

An integer specifying the number of bootstrap samples.

iplot

A logical to indicate whether plots should be generated for visualizing singular value shrinkage.

colCent

A logical indicating whether columns should be centered.

rowCent

A logical indicating whether rows should be centered.

cull

A numeric value for the culling parameter to adjust signal rank.

noisepercentile

A numeric vector specifying the percentiles used for noise estimation for each data block.

noiselvls

A list specifying noise levels for each data block; if NULL, noise levels are estimated internally.

Value

A list containing:

VBars

List of adjusted signal row spaces for each data block.

UBars

List of adjusted signal column spaces for each data block.

phiBars

Vector of adjusted perturbation angles for each data matrix.

psiBars

Vector of loadings perturbation angles for each data matrix.

rBars

Vector of adjusted signal ranks for each data block.

EHats

List of estimated noise matrices for each data block.

singVals

List of singular values before shrinkage for each data block.

singValsHat

List of singular values after shrinkage for each data block.

rSteps

List of signal rank adjustment steps for each data block.

VVHatCacheBars

List of cached VVHat matrices for bootstrap samples.

UUHatCacheBars

List of cached UUHat matrices for bootstrap samples.