## PaGE 5.0 Documentation - the "paired" design

This concept is best illustrated by example. Suppose that two tissue types, type 1 and type 2, are assayed from several different animals. Suppose that a gene is two-fold higher in tissue type 2 than in tissue type 1, but that its intensity level differs greatly from animal to animal. So for example in animal 1 it might be 10 in tissue 1 and 20 in tissue 2, while in animal 2 it is 100 in tissue 1 and 200 in tissue 2. It is a consistently two-fold change in each animal, however this gene might only be found by an approrpriate paired analysis. This is because as unpaired data it has very large variance.

If your data is naturally paired and you suspect an effect of this type, then set this option.

**Note 1**: Both 1-channel and 2-channel data can be paired. But "paired" makes no sense in the case of a

direct comparison design.

**Note 2**: If using the

*t*-statistic option together with the paired option, the 2-sample

*t*-statistic is replaced by the 1-sample

*t*-statistic. If using the

ratio of means option with the paired option, then the ratio of the means is replaced by the geometric mean of the ratios of the paired intensities. See

*t*-statatistic and

ratio of means statistic for more details on the statistics.

**Note 3**: A different permutation scheme is used in the paired case as opposed to the unpaired case. See

permutations for more details on the permutation schemes..