Figure 4.

Precision-recall showing single and multi-block stability selection graphical models applied on simulated data with a block structure. Models are applied on 1,000 simulated datasets (points) with $p=100$ variables following a multivariate Normal distribution corresponding to a random graph ($\nu =0.02$) and with known block structure (50 variables per group, using $v_b=0.2$). The contour lines indicate estimated two-dimensional density distributions. Performances are evaluated in low (a, $n=2p=200$), intermediate (b, $n=p=100$), and high (c, $n=p/2=50$) dimensions.