I am doing a microarray meta-analysis on ~ 3500 samples. all the samples were obtained using the same platform (affymetrix HG_U133PLUS2). Raw data from different studies were downloaded, mixed, then normalized using RMA. I would appreciate if somebody can answer these 2 questions:
1- is it better to do batch effect normalization using raw data or using RMA data?
2- what workflow would you suggest for removing batch effect between studies, so that I may obtain a new data set with batch-filtered intensities?
I would also appreciate if somebody can suggest a workflow for data normalization (intensity and batch effect).
Thank you, Ali Nehme