On Wed, 11 Nov 2009 21:09:57 +0000, Greg Burgess
<[log in to unmask]> wrote:
>Hi SPMers,
>
>I'm looking into running and interpreting PPI analyses, and I have a question.
Rather than
>a PPI analysis on the contrast between two conditions [1 -1 0], can one run
a PPI analysis
>looking at one condition versus baseline [1 0 0]? The former would give you
differences
>in functional connectivity with the seed VOI during Condition A versus
Condition B. If I'm
>interpreting things correctly, I believe that the latter would give you voxels
in which the
>functional connectivity with the seed VOI is greater for Condition A than the
average
>connectivity during the rest of the task (Condition B, Condition C, and
resting blocks). Is
>this interpretation correct?
Sounds right. (A little too early in the morning here... :-) )
>From my quick search for existing studies using PPI analyses, I've only seen
people use
>contrasts of [1 -1]. Is there any reason that other contrasts aren't
commonly used in the
>literature (e.g, [1 0] or [1 -0.5 -0.5])?
If I recall correctly, there's nothing in principle wrong with using the baseline
condition in PPI. The problem is that you have to tell SPM two explicit
conditions. So you have to go back to your "conventional" (non-PPI) analysis
and remodel it with baseline _explicitly_ modeled. See for example this post
by Darren Gitelman, who's one of the PPI gurus:
https://www.jiscmail.ac.uk/cgi-bin/wa.exe?A2=SPM;w6Y61Q;20050503005207-
0500;ind05
Caveat: I tried doing this once, and as a sanity check, I compared
the "conventional" model results with implicit and explicit baselines. They
actually weren't that similar! (Kind of scary in its own right.) I'd recommend
doing that before you proceed.
>Thanks in advance for your responses,
>--Greg
>
>_________________________________________________________________
___
>Greg Burgess, Ph.D.
>Research Associate, Institute of Cognitive Science
>University of Colorado - Boulder
>Email: [log in to unmask]
|