The PaGE confidence measure is based on a permutation algorithm.
The 2-sample unpaired case. When there are two groups and the data is unpaired, and there aren't too many replicates, then PaGE uses all ways of switching the columns of data between the two groups. If the groups have size n and m then there are n+m choose n permutations.
The direct comparison case. In the direct comparison case, unlogged data, the data consist of ratios. A permutation consists of taking a subset of these ratios and forming their reciprocals. If there are n replicates, then there are 2n such permutations. If working with the logged data, a permutation consits of taking a subest of the M values and multiplying them by -1.
The paired case. In the paired case, if unlogged data, pairwise ratios are formed, or if logged data then pairwise differences are formed. The data are thus reduced to the direct comparison case.
Note: If there are more than 2000 permutations then PaGE will just use 2000 random permutations. If you want to raise the default number above 2000 (for greater accuracy), or below 2000 (for greater speed), then use the Configurations dialog box.