Figure 1 : Procedures to generate permutated data.

Let xijk be the smoothed values of CGH data (logR) for the genomic location i=1,2,...,N, stage j=1,2,...,L and samples k=1,2,...,mj
for stage j. Mean value yij of xijk is calculated for each genomic location of each stage. First, the statistics Si is calculated to
investigate difference of yij between different stages: maxyij-minyij where maxyij indicates the largest mean and minyij point
to the smallest mean. Then each difference between two different stages takes into account by Sij=|yi1-yi2| or Sij=yi1-yi2.
The later formula detect in addition to the difference between two stages. To avoid the variability and standardization of
each stage Si is divided by the variance σi for each genomic location. To assess Si (obtained by original data) GISDIP
generates H0 by: 1) Aggregating all samples of all stages and labeling. 2) Permuting the numbers. 3) Picking up the
same amount of samples for each stage and redefine the data as x*ijk. 4) Calculate j*ij by si*. 5) Iterate B times to
generate the H0 by si*(1) ,..., si*(2).

Solvang et al.Journal of Medical Statistics and Informatics  2014 2:10DOI : 10.7243/2053-7662-2-10