Nonlinear Dependence in the Discovery of Differentially Expressed Genes

Figure 2

BRCA example: estimated 𝑝(𝔉,𝐶) moments and estimated distributions using maxent, 𝑝(̂𝔉,𝐶). The distribution estimate on the left uses third-moment information in the maxent optimization, while the right estimate uses only second moments. The third-order estimate exhibits finer details than its second-order counterpart, and a contour that cannot be modeled using a quadratic distribution.