Dear list, has anyone experience with the package "superpc" for developing a gene expression based predictor for survival? We have (Affymetrix) arrays from two different types which we normalized separately and which consequently have different means and variation. We would like to use one set for training and the other for testing. My question is whether it is necessary to bring the test set first on the same scale as the training set. I have had a look into the code of superpc and I have seen that there is done some centering and scaling, so eventually no previous steps have to be performed to make train and test data comparable? Thank you very much Best regards Manuela -- -------------------------------------------------- Dipl.-Stat. Manuela Hummel wissenschaftliche Mitarbeiterin IBE - Institut f?r med. Informationsverarbeitung, Biometrie und Epidemiologie Ludwig-Maximilians-Universit?t M?nchen Tel: +49 89 7095 7497 mailto: hummel at ibe.med.uni-muenchen.de URL: http://ibe.web.med.uni-muenchen.de