Fig. 4.

Simulation results. (A)Mean size of significant clusters at cluster-level p < .05 FWER corrected. The horizontal dashed lines indicate the average cluster sizes of anatomical regions of the Harvard–Oxford atlas (>50% probability) and functional parcels of Craddock’s 200 parcellation solution (CR200; Craddock et al., 2012). (B)We present (a) the estimated voxel-level false discovery rate ( $\widehat{\text{vFDR}}$), (b) the estimated voxel-level sensitivity ( $\widehat{\text{vSens}}$), and (c) the estimated cluster-level family-wise error rate ( $\widehat{\text{cFWER}}$) from thresholded brain maps at cluster p < .05 FWER corrected. For detailed simulation procedures and definitions of the evaluation measures, please refer to Fig. 3, Table 3, and Simulation 2 section. The horizontal dashed line for the ( $\widehat{\text{cFWER}}$) plot indicates the level that the cluster-extent based thresholding procedure controls. The error bars represent standard error of the mean (S.E.M.). If the error bars are smaller than the markers, they would not be visible. The estimated values are represented with a hat. ^aSNR = signal-to-noise ratio, which indicates the true signal mean divided by the standard deviation of the noise.