3dfim+ gives many zeros

AFNI Message Board
1

Hi AFNI experts,

When I use 3fim+ for the seed-based analysis, I often see many zeros (many small “holes” in the final correlation map). I had always thought that they were real zeros. But I just noticed that sometimes, it happens even very close to the seed, (and all neighboring voxels are close to 1, as they should).

This is my command line (mask_ideal is created with 3dmaskave in a mask created with 3dUndump). I tried with different -out options, but these zeros are always in the same voxels.

3dfim+ -input errts_file -ideal_file targetTimeSerie -out 'Spearman CC' -bucket $targetCorrelationMap

I took the timeserie for one voxel at the time, and computed the correlation coefficient:

     3dmaskave -q -xbox x y z errts_file >> Voxel_x_y_z_timeSerie.1D
     3dfim+ -input1D Voxel_x_y_z_timeSerie.1D -ideal_file targetTimeSerie -out 'Spearman CC'

In non-zero voxels, the result was exactly the same. However, in the buggy voxel, I suddenly had a more realistic result (close to 1).

Can you please advise me on what to do?

2

I have seen that program lose results that
have too high of a statistic before. But we
do not maintain it anymore, it is just there
for people with existing processing streams
that they want to maintain.

For this case, use 3dTcorr1D, as in:

3dTcorr1D -spearman -prefix$target.spearman errts_file targetTimeSeries.1D
3dTcorr1D -pearson -prefix$target.pearson errts_file targetTimeSeries.1D
  • rick
3

Oh! Thank you!