Figure 2.

Diagram for further sparsity control using $\epsilon$. $\epsilon =\tau {\left|\right|y\left|\right|}_2$; the entries of $\delta$ were sorted in descending order by amplitude. The results were generated with white noise simulation data set of size 100 × 1e⁶ (50 cases and 50 controls) with $k=0.02\times \mathsf{\text{1}}{\mathsf{\text{e}}}^{\mathsf{\text{6}}}=\mathsf{\text{2}}{\mathsf{\text{e}}}^{\mathsf{\text{4}}}$