Hi Woogul,

In your 3dcalc command to scale the data, dataset -c should be a
mask (or omitted). You are using Run_all_mean+orig, which does
not look right. If that is an additional scaling, it could cause
an integer overflow artifact if you are working with short integer
datasets.

What -polort option did you give to 3dDeconvolve? Either each run
has a constant baseline, or 3dDeconvolve does not know you have 3
runs (in which case you need to add the -concat option).

That makes sense for regressor C. Right, it is not the baseline.

What does 3dinfo say about Run1_3dcalc+orig?

---

So these are group results then. It's hard to say. You should
be sure of the single subject results first. Do they look good?
Do they look like what you would expect? Do you see appropriate
contrasts over the ROIs for some individual subjects?

Drawing ROI regions is very time consuming, added on to the fact
that it produces ROIs that vary across subjects. That is why
people usually get them from an atlas. Getting them from an atlas
also makes the regions independent of the statistical computations,
which is good.

If you make functional ROIs at the group level, that may work.
But then you must be careful not to make any additional statistical
claims in those ROIs, as you have already selected "active" voxels.

- rick