Hi Joe,
Each iresp curve shows the consistent response to that particular
stimulus (per voxel). I find those curves terribly useful, as they
show what is really in your data. Even if you publish data that
assumes GAM, it is good to analyze some of that data using TENTs,
just to verify that the responses really do look like GAM curves.
Usually, the main reason for using TENTs is that you don't want to
make any assumptions about the shape of the response curve (other
than the maximum response time), because it is expected that the
curves will vary across some brain regions. Contrast that with
using GAM as a basis function, where you assume it looks _exactly_
like that curve, and the only question to answer is "how big was
the response?".
In your case, I guess you expect the shape of the response curves
to vary (e.g. in terms of how soon the response occurs, or how long
it lasts). If so, it seems quite likely that you would want to
discuss how average response curves vary over ROIs. Running either
3dmaskave or the more general 3dROIstats on your iresp curves seems
quite appropriate.
If you would like to read more about "deconvolution using TENTs",
check out the class handout on it. It is under the full handout
directory (/pub/dist/edu/latest):
[
afni.nimh.nih.gov]
[
afni.nimh.nih.gov]
As usual, there is a .ppt file, along with .pdf and .html formats.
- rick