Evaluation of Second-Level Inference in fMRI Analysis

Figure 4

ROC for the low signal strength (SNR = 1) and for the high signal strength (SNR = 2.5) with identical subject-specific variability; for permutation-based inference and for parametric inference for cluster-based inference with : cluster-FWE: familywise error correction based on cluster-size inference, cluster-FDR: False Discovery Rate correction based on cluster-size inference, and BCL: two-step procedure with a Bonferroni-like first threshold and minimal cluster size of 10. OLS: Ordinary Least Squares approach and WLS: Weighted Least Squares approach.