Hello all, I'm wondering if someone could give me some information about the best way to proceed with my analysis.

I have 9 groups and 4-time points, and each group has 3 samples. I'm wondering what is the best strategy statistically speaking. Should I do RMA for all CEL files at once? or do it per time point? RMA is the only normalization I should perform?

Then I'm wondering the same thing after normalization. To run Limma, should I use all the information in my design matrix? Let's say I do not want to compare the time course, should I then only take the information of 1-time point to run limma?

Any advice would be appreciated it.

Ivan

Unless you expect really big differences (in variability, mostly) between the groups, you should normalize together and analyze together. Even if you only compare two groups, having the other groups in there helps estimate variance.