Dear List, I would like to ask a question about using limma package to compare multiple groups. As I understand, if I use multiple testing method "separate", I will not have a p-value correction for testing multiple contrasts, since this method does multiple testing for each contrast separately. If I use "method = global", all tests are combined into a long vector of tests, and therefore all tests are treated as equivalent regardles whether they relate to probe or contrast. For the other two methods (hierarchical and nestedF), their properties are not well understood with adjustment "BH". Therefore, I was planning to use F.p.value (moderated F-statistic), and to correct it for multiple testing with BH. However, this will only provide me with one p-value related to all pair-wise comparisons (all contrasts) for one probe set, and won't point out in which pair-wise comparisons exactly this probe set is differentially expressed. If I were to use ANOVA analysis, I would've used a post-hoc procedure in order to find an answer to this question, however I am not sure how to proceed with the methods offered by limma package. Can anybody, please, explain to me how can I perform a procedure similar in its meaning to post-hoc test in order to find exactly between which groups are the probe sets differentially expressed? Thank you very much! Anna Dvorkin-Gheva [[alternative HTML version deleted]]