corBootstrap: corBootstrap

In ProCoNA: Protein co-expression network analysis (ProCoNA).

Description

Boostraps a correlation matrix. In order to bootstrap a large correlation matrix, several thousand samplings may be necessary. To avoid storing thousands of matrices, a running mean is kept for each pairwise correlation. In addition, a running standard deviation is computed so that for each pairwise correlation, we can estimate the distribution of values across resamplings. After each resampling, a new correlation matrix is computed. A difference is taken between this new matrix and the running mean. If all differences are less than the specified threshold, then the bootstrapped matrix has converged to a final state.

Usage

corBootstrap(dataMatrix, networkType = "signed", threshold = 1e-04, 
    tmpSaveFile = TRUE)

Arguments

dataMatrix	Matrix with samples in rows and peptides (or other data type) in columns.
networkType	Whether the sign is considered in constructing adjacency and TOM
threshold	Maximum difference allowed between running mean boostrap correlation matrix, and new resampled cor matrix. Defines how soon we consider the bootstrap to have converged.
tmpSaveFile	Should temporary saves be done?

Value

Returns a list of the bootstrapped matrix, standard deviation matrix, and the number of resamplings done.

Author(s)

David L Gibbs

Examples

data(ProCoNA_Data)
x <- peptideData[,1:10]
y <- corBootstrap(dataMatrix=x, networkType="unsigned", threshold=0.1, tmpSaveFile=FALSE)

ProCoNA documentation built on Nov. 1, 2018, 3:28 a.m.