I am reading the paper of Bolstad 'A comparison of normalization methods for high xensity oligonucleotide array data based on variance and bias' and was confused by the definition of the baseline array.

...to choose a baseline array, in particular, the array having the the median of the median intensities.

So, from all chips I choose one chip, to which all others with be normalized. And this specific chip should have the median of the median intensities. It seems to be a bit artifical, because in order to construct an array with the median of the median intensities I need to iterate over all probes and find the median and write it into the array and I do it for every probe. So, I do not choose a chip from my collection of the chips but I create an artificial array with the median of tbe median intensities. Do I understand it right? Or i pick up the chip that has the best intensities to fit the definiton 'median of the median intensities' ? It is not clear for me if I either pick up a chip with the 'median of the median intensities' characteristics or I construct this array artificially.

Thanks in advance