Fig. 3
From: Learning the optimal scale for GWAS through hierarchical SNP aggregation
Recall vs Precision for each method (shape and colors in plot). In rows are the simulation scenarios. In columns, we evaluate performance using Benjamini-Hochberg threshold (left) and bonferroni correction threshold (right). The second row illustrates the performance to retrieve the true causal genomic region under the SNPclus scenario, thus only group-based approaches are considered (SASA, SKAT.tree and SKAT.notree). The numbers inside the points correspond to the number of causal predictors and each point is the average value of 5 replicates