Figure 1
From: Kerfdr: a semi-parametric kernel-based approach to local false discovery rate estimation
Semi-supervised. Root Mean Square Error (RMSE) between the true local FDR τ and the estimates as a function of the proportion 1 - π0 (log-log scale). Proportion of known hypothesis: κ = 0 (dotted), 1% (cross), 5% (asterix), 10% (circle) and 50% (square). Top: exponential shape for f1. Bottom: uniform shape. Left: μ = 0.001. Right: μ = 0.01. Variance of the RMSE lies between 1e-4 and 5e-4 with 500 simulations.