Hello Jay:
I set the maximum % Change value for 3dfim+ at 1000. Since, for BOLD signals,
the true % Change should be only a few percent, the limit of 1000 seems
reasonable. Anything that large or larger is most very likely extraneous.
The reason for limiting the max % Change value is that the data are stored
as scaled short integers; hence, the accuracy of the stored data is limited
by the largest value for any voxel in the volume for that sub-brick.
You can use the "-input1D" option of 3dfim+ to examine the results for an
individual voxel, in order to validate the result. Either dump the time series
for your favorite voxel, or you can make up your own "time series data".
Suppose that you write the fMRI data for voxel #042_033_001 to a ".1D" file.
Then the command line:
3dfim+ \
-input1D 042_033_001.1D \
-nfirst 5 \
-polort 1 \
-ideal_file myIdeal.1D \
-out All
will write the results to the screen for that voxel.
Or, at least that's the way it used to work (I seem to be using that phrase
a lot recently). It appears that recent changes (attn: RWC) have disabled
the ability of 3dfim+ to process a single time series.
Alternatively, you could use program 3dDeconvolve:
3dDeconvolve \
-input1D 042_033_001.1D \
-nfirst 5 \
-polort 1 \
-num_stimts 1 \
-stim_file 1 myIdeal.1D -stim_label 1 "myIdeal" -stim_maxlag 1 5 \
-fitts myFit
1dplot -ignore 5 -one 042_033_001.1D myFit.1D
Doug Ward