Hi, Am I correct in saying that the default setting of vsn2 is to scale everything to the first array? Could this be potentially problematic? For example, we have observed strange results when we change the order of loading of the data into vsn2. The resulting normalised data looks different when vieved in an MA plot. Ros

Dear Roslin, > Am I correct in saying that the default setting of vsn2 is to scale > everything to the first array? This is not the case. The model that vsn fits is invariant under permutation of the arrays (you can look at the vignette "Likelihood Calculations for vsn" if you are interested in details). However, the fitting does includes an iterative numeric optimisation; if that fails, and runs into a local optimum, such a behaviour can depend sensitively on the data (not just the order of the arrays). > Could this be potentially problematic? For example, we have observed > strange results when we change the order of loading of the data into > vsn2. The resulting normalised data looks different when vieved in an MA > plot. Can you send me an example that I can reproduce? Such examples would be helpful for me to improve the diagnostics that can be done by vsn2, either automatically, or in a subsequent quality control step. Best wishes Wolfgang ------------------------------------------------------------------ Wolfgang Huber EBI/EMBL Cambridge UK http://www.ebi.ac.uk/huber