The test statistic for differential modular structures in two networks is sensitive to the choice of &epsilon, so we tried the range of values listed below. The following table shows the results for various choices of &epsilon and also presents the value of the test statistic N, its corresponding p-value, the number of modules discovered in each network, and the total number of genes discovered in some module for each network.
&epsilon
N
p-value
# of modules in ever-CFS network
# of modules in non-fatigues network
.35
.174
.211
1
1
.40
.286
.421
1
1
.45
.586
.385
1
1
.50
.978
.153
7
1
.55
.997
.098
5
1
As seen in the table, the value of the test statistic and the structure of the networks are sensitive to the selection of &epsilon . However, the result of the significance test is not affected by our choice of &epsilon for this analysis. Thus, regardless of how we define a module, observed data does not support the statement that the two networks have significantly different modular structures.