In trying to automate a few tasks, I am writing a script that does the same thing as "Compute Fim+" from within AFNI. I see that 3dfim+ from the command line should give nearly the same result (within rounding error) as using AFNI.
Fit Coef & Correlation values seem comparable using either method, slightly different but within rounding error.
% Change seems dubious:
1) With % Change values, somthing is strange: in the "Define Function" pane, the "ranges of dataset values" shows up as -1000 to 1000 with a dataset created using 3dfim+, -506.0132 to 2523.673 with AFNI (using the same input dataset). Using "Func Info", I found the following information:
3dfim+:
-- At sub-brick #1 '% Change' datum type is short: -32767 to 32767 [internal]
[* 0.0305185] -1000 to 1000 [scaled]
AFNI:
-- At sub-brick #1 '% Change' datum type is short: -6570 to 32767 [internal]
[* 0.0770188] -506.013 to 2523.67 [scaled]
2) When all three image windows are open, the Func values can be displayed in the "Define Function" pane. There are several voxels that are "0" using the AFNI method and a negative number (e.g. -70, -300, -777) using 3dfim+.
I'd like to make sure that these methods produce the same results. Any thoughts?
Thanks in advance,
- Jay