Hello Frank:

Yes, something did change.

Program 3dFDR reads in the bucket dataset. Non-statistical sub-bricks are
copied directly to the output. Only statistical sub-bricks are converted
to the corresponding FDR z-scores. At least, that's the way it used to work.
But not anymore.

The gory details: (WARNING: This story is not for the fainthearted.)

To determine if a sub-brick is a statistical sub-brick, program 3dFDR
calls function FUNC_IS_STAT. Now, FUNC_IS_STAT in turn calls FUNC_HAVE_PVAL
to determine if the sub-brick has a corresponding p-value. But, FUNC_HAVE_PVAL
calls FUNC_HAVE_THR to determine if the sub-brick has a threshold value.
Now, here is where the story gets interesting. Prior to 06 Feb 2003,
fim type sub-bricks were coded as NOT being threshold type sub-bricks. However,
on that date, someone (with the initials RWC) changed the coding for the
fim type sub-bricks. So, now the fim type sub-bricks can be thresholded.
But this has the unfortunate side effect that fim type sub-bricks are now
identified by the software as being statistical type sub-bricks, and so
they are transformed to FDR z-scores (which are meaningless).

Doug Ward