Hi, I just recently saw R-function to compute Q-values given unadjusted p-values on John Storey web page: http://www.stat.berkeley.edu/~storey/qvalue/index.html I have not tried it myself. My question concerns normalization discussion on which I have been reading in recent postings. Does anyone have suggestion of what would work for 2-color arrays with common reference design (and the "truth" for the reference is known -- special case of measuring copy number rather than expression) where majority of the genome is rearranged (i.e. majority of the spots "change"). Assumption of equal fraction of "ups" and "downs" may not be made either. Generally simple global median normalization works just fine but not here. I have been hitting my head off the the wall for the past 3 weeks after encountering three datasets in the row with significant proportion of tumors looking like normalization failed because of global genomic instability (majority of clones show change) but can't come up with a good normalization idea here rather than change the post-normalization analysis approach. No internal controls are possible here. Has anybody encountered situation like that with 2-color gene expression chips? I have 3 replicates for each spot but they are located right next to each other. Thanks you very much Jane