Hi Todd,
The results for the regressors A+B and A-B are just the main effects of your task - either areas that are active in both conditions A and B, or in condition A more than condition B, respectively. The only difference from if you had put these regressors into a non-PPI analysis is that the results you are seeing are for the variance which is also not explained by your seed ROI timecourse, or by the PPI. But I imagine if you were to do a straightforward analysis modelling your data wiith A+B and A-B (without any PPI regressors), or indeed modelling A and B separately and comparing in a 2nd level contrast, you would probably see similar results.
Jill
-----------
Dr Jill O'Reilly
FMRIB Centre, Nuffield Dept Clinical Neurosciences, Oxford University
--------------------------------------
From: Todd Thompson <[log in to unmask]<mailto:[log in to unmask]>>
Date: 3 August 2011 07:16:14 GMT+01:00
To: [log in to unmask]<mailto:[log in to unmask]>
Subject: [FSL] Help interpreting B-A PPI model results?
Reply-To: FSL - FMRIB's Software Library <[log in to unmask]<mailto:[log in to unmask]>>
Hi, all.
I have a (conceptually) simple experiment where subjects try to
regulate auditory attention. In up-blocks, they try to increase
activation in a pre-defined ROI; in down-blocks they try to decrease
activation in that ROI. (This was a real-time fMRI experiment, and I'm
now analyzing the data offline.) I'm trying to ask the question: "When
people are successfully self-regulating, where are the 'control'
regions?"
I hypothesize that the "up" control regions may differ from the "down"
control regions, so I run a PPI model as suggested here:
http://www.fmrib.ox.ac.uk/Members/joreilly/frequently-asked-questions-ppi
For each subject, my four regressors are:
1) Up-Down, convolved with the HRF and filtered
2) The timecourse from the functionally localized ROI
3) The PPI regressor
4) Up+Down, convolved with the HRF and filtered, in order to "span the
A+B vector space" as mentioned on the page above
These are then passed up into a group model.
I know that my primary regressor of interest is the PPI regressor,
because that tells me areas that are more connected with my ROI in up
vs. down. The physiological regressor is relatively boring -- it
correlates with the whole brain. But the other two maps look really
interesting (and significant), and I can't quite figure out how to
interpret them.
What does it mean to talk about an "Up+Down" or "Up-Down" contrast in
a model where you've already included the other three regressors?
Thanks. I'm sure there's an intuitive way of thinking about this, but
I'm getting stuck on it...
Todd
|