[VoxBo] group F-test failure and FDR

[Daniel Y Kimberg]

Hi all.  I wanted to get out a quick post now that I've resolved the
issue David January posted about earlier, involving some F maps gone
awry.

Part of the issue was a bug in vbstatmap that made it occasionally
generate all-NaN maps.  Another part of the bug was in the handling of
old-style contrasts.  In short, the following wasn't always working:
  -c "1 1 1" -s f
but the following was:
  -c "my_f_contrast f vec 1 1 1"

The second way is the way you specify contrasts in the contrasts.txt
file, and it's strongly recommended to avoid the old style.  Just to
avoid issues, I will probably make vbstatmap stop supporting the old
style soon, so existing scripts should be converted if needed.  In any
case, I've uploaded new versions of vbstatmap and vbdumpstats to the
ftp server, and they can be retrieved from the following links, for
reasonably modern Linux distributions.

ftp://voxbo.org/pub/voxbo/alacarte/linux-bleeding/vbdumpstats
ftp://voxbo.org/pub/voxbo/alacarte/linux-bleeding/vbstatmap

Users at the Penn CfN already have the new stuff.  I'll post non-Linux
or old-Linux versions if anyone asks.  Please let me know if these
give you any trouble.

dan