Dear Xiuqin Jia,
> When Canonical HRF with time and dispersion derivatives is used for
> within-subject analysis with SPM5, how can I get the best contrast vector
> for the t-test between the task versus baseline condition? Should it be [1 0
> 0 -1 0 0] or [1 1 1 -1 -1 -1] or other better choice?
Including derivatives in addition the the canonical HRF will
definitely help explain your data better (which is generally a good
thing), but it can also make the interpretation more difficult. In
your case:
[1 0 0 -1 0 0]
would give you the difference in canonical HRF across conditions,
which is the most straightforward to interpret, and is generally ok to
interpret as a difference in the magnitude of response across these
two conditions.
A t contrast of:
[0 1 0 0 -1 0]
would tell you where the two conditions differed in their temporal
derivatives, which without other information is not very helpful.
A t contrast of:
[1 1 1 -1 -1 -1]
tells you essentially where [the average of condition1 canonical HRF,
temporal derivative, and dispersion derivative] differs from [the
average of condition2 canonical HRF, temporal derivative, and
dispersion derivative]. This is essentially meaningless and is not
what you want.
If you want to test more than just the canonical HRF, then an F test
is the way to go, where each row tests a different regressor:
1 0 0 -1 0 0
0 1 0 0 -1 0
0 0 1 0 0 -1
However, this is also difficult to interpret, because even if you get
a significant result, it won't be clear whether this is due to the
canonical or one of the derivatives. In addition, because an F test
is not signed, you won't know whether it is due to
condition1>condition2, or condition2>condition1. So, generally to
interpret significant F statistics it's necessary to extract the data
and plot it.
This is a long way round of saying, this is a common issue that comes
up and there isn't a perfect solution. The simplest solution is to
use a t-test over just the canonical HRFs, so [1 0 0 -1 0 0] in your
case.
There are a number of previous posts on this topic that may be
helpful; if you try searching for "temporal derivative contrast", for
example, these include:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;2369f6db.1102
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;e9870ebf.1011
https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=SPM;7fba87b3.0811
Hope this helps!
Best regards,
Jonathan
--
Dr. Jonathan Peelle
Department of Neurology
University of Pennsylvania
3 West Gates
3400 Spruce Street
Philadelphia, PA 19104
USA
http://jonathanpeelle.net/
|