Sean, thank you for the answer!:)
One more quick question: in this fresh draft by Kasper D. Hansen they have analysed 450k data. For 2 datasets they have downloaded IDAT files and then normalised them. For one dataset they didn't apply normalisation, because only methylated, unmethylated intensities were available (Material and Methods - Processing of the methylation data).
I guess that K. Hansen knows what he does, but will this work for every dataset? For example: If I download only methylated/unmethylated values can I skip normalisation and go on further?