For (q>1), the distribution function exhibits a thermal cutoff on the maximum value allowed for the velocity of the ions, given by begin{aligned} v_mathrm{max}=sqrt{frac{2T_i}{m_i q-1)}}, end{aligned}beyond which no probable states exist.
We therefore introduce the use of a flexible cutoff C, which is a percentage that is used for determine the cutoff on the test values (i.e. p-values or the difference in relative frequency), and is based on the number of detected genes from a comparison and their corresponding test values.
For fuzzy clustering methods, we obtained crisp but potentially overlapping modules by placing a cutoff on the membership values.
GO term statistics were obtained via topGO [ 34] with a 5% cutoff on the P-values ("weight.log" method).
The decision trees also showed that the size of the libraries is of importance when setting a proper cutoff on the test values, since this attribute appeared as a node in the trees.
In addition, the simulation studies show that the combination of AC, Fisher's and Diff increases the number of detected up-regulated genes, and that the size of the sets is important when setting a proper cutoff on the test values.
While in q > 1, the distribution function has a cutoff on the maximum permitted value for velocity of the particles, given by ( v_{hbox{max} } = sqrt {2 /left( {q - 1} right)} v_{T} ).
In the simulation studies we manually investigated the generated statistical values from the comparisons in order to set proper cutoffs on the test values.
Initially, different cutoffs on the test values were implemented and tested (Table 2, rule R1-R22).
For different possible cutoffs on the test values, e.g. p = {0.001, 0.005, 0.01, 0.05, 0.1}, the percentage number of genes with ≤ p is calculated.
