[VoxBo] filtering .tes files
Daniel Y Kimberg
kimberg at mail.med.upenn.edu
Wed Oct 4 21:21:12 EDT 2006
Charan Ranganath wrote:
> Sorry Dan, what I meant is that the voxel surfer and VBview can spit
> out a filtered, adjusted time series from an individual voxel. What
> I wanted was to create a new .tes file with the filtered and
> adjusted timeseries from each voxel. Can you do that with
> vbmakeresid?
Sure -- the only difference between residuals in the usual sense and
what you're asking for is that you only want to remove the covariates
of no interest. So you can just create a model with only those
covariates, and the residuals will be what you need.
When you run a glm in VoxBo, a residual volume is created at the same
time as the regression (since the resids are on hand right then),
called myglmname.res. Unfortunately, it's hardcoded to save only 64
time points (volumes), evenly spaced through your data. The reason
for that is that it's meant to be used to estimate spatial smoothness,
and the full residual volume would be needlessly large (same dims as
your full 4D data, 64 bits per voxel). So you would have to use
vbmakeresid for this. If you omit the -n flag, it will do all time
points.
Looking at the code for vbmakeresid, I'm concerned that vbmakeresid
might not handle detrending and mean scaling, so I will double-check
that. I'm back coding for the next few days, so I should be able to
squash bugs (including your favorite, the one with loading a second
GLM in vbview), and maybe see if I can make vbmakeresid a little
faster.
dan
More information about the voxbo-general
mailing list