Sorry, I was unclear. I want to input the peaks of my conditions (voxel-wise) into 3dAnova. Using the -max option in 3dTstat to calculate the maximal point within the IRF yields the largest value (i.e., most positive). However, when looking at deactivations, the data point of interest is the most negative value. For deactivations, the most positive value is the closest to baseline and belies the impact of the task.
One way around this is to use -mean in 3dTstat to calculate the mean of a few points where the peak (or trough) should be. This is probably what we'll do. I just wondered if it was possible to extract the most deviated point from fixation baseline (whether that point was positive or negative). I want to end up with a full F activation map that reflects all modulation by task conditions whether that modulation is in a positive or negative direction relative to fixation baseline.
Does this make any more sense?