hackedSVA: Fast Surrogate Variable Analysis

In MalteThodberg/BEEF: Batch Effect Estimation with fSVA (Frozen Surrogate Variable Analysis)

Description

Fast Surrogate Variable Analysis

Usage

hackedSVA(dat, mod, mod0 = NULL, n.sv, B = 100, alpha = 0.25,
  epsilon = 0.001, VERBOSE = F)

Arguments

dat	the measurement matrix, where rows are features and columns are samples.
mod	the model matrix being used to fit the data.
mod0	the null model matrix.
n.sv	the number of surrogate variables to estimate. The use of random matrix theory is recommended to estimate n.sv. See the example for more details.
B	the maximum iteration number.
alpha	determines the initial point for optimization which affects the convergence rate.
epsilon	the convergence threshold. The Spearman's correlation between posterior probabilities of consecutive iterations of the algorithm is compared to epsilon. Empirical evaluation on several data sets revealed epsilon=0.005 gives very reasonable results. However, we suggest epsilon=1e-3 as a conservative threshold.
VERBOSE	a logical variable. If TRUE, prints some details about iterative progress of the algorithm.

Value

Returns a list containing the surrogate variables and some meta data about the convergence criterion.

MalteThodberg/BEEF documentation built on May 7, 2019, 2:09 p.m.