[BioC] ComBat: Comparison of data

[W. Evan Johnson]

Peter, 

Use hierarchical clustering for before and after to see if the samples cluster by batch. Also many people like to plot the principle components to see if there is separation by batch in some of the components. Finally, if you have only a couple batches, you could try to predict batch based on the data and see if this is different between the corrected and uncorrected data.

Thanks!

Evan



On Oct 29, 2012, at 6:35 AM, Peter Kupfer wrote:

> Dear all,
> i deal with the tool ComBat and try to remove underlying batch effects. Therefor I used covariates and recommend in the manual. 
> After correcting the data and want to have a glue, what the tool did. Therefor i am looking for a method for comparing the corrected and the uncorrected data. 
> Is there any method for such comparison which you would suggest? Furthermore I am looking for the R code for the (so called) shrinkage plots. 
> Would be amazing if anybody could give me some ideas to compare the outcome of ComBat with my input.
> Thanks in advance and all the best from Germany
> Peter Kupfer
> 
> 
> 
> 
>