DoISVA
(Package: isva) :
Feature selection using independent surrogate variables
Given a data matrix and a phenotype of interest, this function performs feature selection to identify features associated with the phenotype of interest in the presence of potential confounding factors. The algorithm first finds the variation in the data matrix not associated with the phenotype of interest (using a linear model), and subsequently performs Independent Component Analysis (ICA) on this residual variation matrix. The number of independent components to be inferred can be prespecified or estimated using Random Matrix Theory. Independent Surrogate Variables (ISVs) are constructed from the independent components and provide estimates of the effect of confounders on the data. If potential confounders are unknown (default NULL option) there will be as many ISVs as there are independent components in the residual variation space. If potential confounders are known (either exactly or subject to error/uncertainty) the algorithm will select only those independent components that correlate with the confounders. If potential confounders are specified it can happen that ISVA will not select any ISVs because none of the independent components correlates with the confounders. In this scenario ISVA should be rerun with the default (NULL) option. The constructed ISVs are finally included as covariates in a multivariate regression model to identify features that correlate with the phenotype of interest independently of the potential confounders.
Independent Surrogate Variable Analysis is an algorithm for feature selection in the presence of potential confounding factors, specially designed for the analysis of large-scale high-dimensional quantitative genomic data (e.g microarrays). It uses Independent Component Analysis (ICA) to model the confounding factors as independent surrogate variables (ISVs). These ISVs are included as covariates in a multivariate regression model to subsequently identify features that correlate with a phenotype of interest independently of these confounders. The ICA implementation used is that of the fastICA R-package.
isvaFn
(Package: isva) :
Main engine function for inference of independent surrogate variables (ISVs)
This is the main engine function which infers the statistically independent surrogate variables (ISVs) by performing Independent Component Analysis (ICA) on the residual variation matrix. It uses the ICA implementation of the fastICA R-package. The residual variation matrix reflects the variation orthogonal to that of a phenotype of interest and is inferred using a linear model.
EstDimRMT
(Package: isva) :
Estimates dimensionality of a data set using Random Matrix Theory
Given the data matrix, it estimates the number of significant components of variation by comparing the observed distribution of spectral eigenvalues to the theoretical one under a Gaussian Orthogonal Ensemble (GOE). Specifically, a spectral decomposition of the data covariance matrix is performed and the number of eigenvalues larger than the theoretical maximum predicted by the GOE is taken as an estimate of the number of significant components.
providing 
.