Johnson and Li. 2007, Adjusting batch effects in microarray expression data using empirical Bayes methods, is the paper for the comBat package, and I have trouble understanding the following from page 119 

..... stabilize gene variances by shrinking variances across all other genes, possibly protecting their inference from artifacts in the data. In this paper, we extend the EB methods to the problem of adjusting for batch effects in microarray data.  Could anyone please explain "shrinking variances across all other genes"

Thanks!

This support site is intended primarily as a source for people having technical problems with Bioconductor software, not as a source of primers on statistical methodology. You could instead google something like 'empirical bayes variance shrinkage' - I am sure there is some approachable explanation on the internet.