Hello AFNI experts,
After running tshift during afni_proc, some voxels within the brain become negatively valued. These voxels are in areas with low signal (MTL and ventral PFC). I was wondering if this is typical for data with poor SNR?
This would just be due to the interpolation applied, and are usually due to a large dip toward zero.
For example, assume the values go 50 1 80. Since the tshift operation is interpolating off the temporal grid, a new value might come close to the time of the ‘1’, but slightly before or after, for example.
The key point is that the interpolation will almost certainly not have it’s minimum right at that ‘1’, and in such a case, it would almost certainly go negative. Then it is just a question of whether the shifted time point hits where the resampled function goes negative.
Does that seem reasonable?