Batch effects are removed using a parametric empirical Bayes approach, as described in Johnson et al. 2007.
Usage
combat_p(Y, b, covariates = NULL, prior.plots=T)
Arguments
Y
A matrix with $n$ rows and $g$ columns, where $n$ is the number of objects and $g$
is the number of variables. In the case of gene expression data, columns correspond
to genes (probe sets) and rows to samples.
b
A vector of class factor with the element in position $i$ ($i=1,\ldots,n$) representing
the batch from which observation $i$ belongs to.
covariates
An object of class data.frame where each column corresponds to a quantitative
variable (of class numeric) or a qualitative variable (of class factor).
prior.plots
A logical value; if true prior plots to compare kernel density estimate and parametric estimate are produced.
Value
A matrix of adjusted data with $n$ rows and $g$ columns.
Details
In this implementation pre-processing of the data is obtained through Moore-Penrose pseudo inverse.
Therefore the final matrix of adjusted data can be different from that obtained through the
original ComBat algorithm (see sva package). NA values are not allowed.
References
Johnson et al. 2007 http://www.ncbi.nlm.nih.gov/pubmed/16632515