Figure 5:

Case A: roughly compatible statistical evidence, but with important differences in effect estimates. Group differences are shown for the incongruent condition at (X,Y,Z) = (39, 29, 56) in the left postcentral gyrus. (top row) All three approaches showed strong statistical evidence for the presence of a group difference. (bottom row) The effect estimation of group difference differs importantly across methods. In each panel, vertical bars and shaded bands indicate standard error. As the estimated HRFs are largely similar to the canonical version, the estimated group contrast (0.15±0.03) from canonical HRF (first column) is roughly compatible with the HRF peak differences from the sampled and smooth HRF estimates. However, as the canonical HRF assumes a peak at 5 s, slightly later than what was revealed through HRF estimation, it substantially underestimated effects for both groups. Anticipation effects occurred about one TR before stimulus onset as illustrated by the HRF estimation approaches. HRFs for the other condition (BP con, and HV con) and neighboring voxels shared similar profiles (not shown here).