Figure 1.

Smoothing artifact of the center voxel for different analysis methods as a function of the configuration size and CNR for simulated motor activation data. Note that the figure is composed of 9 separate images belonging to configuration sizes 1 to 9, and each image for a particular configuration size is a function of the CNR ranging from 0 to 1 in steps of 0.1 in the horizontal direction. The center voxel for each configuration is by design nonactive and the other members (neighboring voxels belonging to the particular configuration) are active with given CNR. In general the smoothing artifact increases with increasing CNR except for single-voxel analysis, where no artifact exists (as expected). Regarding the computation of the false positive fraction, the statistical thresholds were chosen corresponding to a family-wise error rate (FWE) < 0.05 for re-sampled restingstate data.