Hi Vladimir,
Thank you for the information. If I may ask, what kind of
statistics are you referring to with respect to assessing significance of
winning model parameters between sessions? I guess more specifically, is it not
a result in itself if two different models are selected as the winning model in
the two sessions (if both models were included as potential options in both
sessions)? If I can, I think I would prefer being able to run the data
separately, but would like to be able to make inferences about changes in
network activity between the sessions within one experiment as well as between
experiments.
I am creating a head model for source reconstruction for each
session, which I think should resolve the issue of head position changes between
sessions, correct?
Thank you for the LFP suggestion, because I was also curious as
to what would be the best neuronal model to use in this paradigm.
Best,
Vivek
From: Vladimir Litvak
[mailto:[log in to unmask]]
Sent: Wednesday, January 25, 2012 5:00 PM
To: Buch, Vivek (NIH/NINDS) [G]
Cc: [log in to unmask]
Subject: Re: [SPM] DCM analysis set-up
Dear Vivek,
Both are possible but I think the second (concatenating)
approach is more common. You could also find one winning model for all
subjects/sessions fitted separately and then do classical stats on the
parameters of that model. Since the head position is likely to differ between
the two experiments I'd suggest that you use some kind of source extraction
(e.g. beamformer) and then LFP option in DCM.
Vladimir
On Tue, Jan 24, 2012 at 10:01 PM, Buch, Vivek (NIH/NINDS)
[G] <[log in to unmask]> wrote:
Hi
SPM experts,
I
have a conceptual question regarding the best way to set up some DCM analysis.
Subject
Experiment
1:
Four
separate MEG sessions
Experiment
2:
Four
separate MEG sessions
As
above, basically I have two experiments with four MEG sessions each. I would
like to compare how network activity changes across the different sessions
within experiment and then compare those results between experiments. There is
no event related modulation during any of the scans. Experimental influence is
occurring between sessions, outside of the scanner, and is different for
Experiment 1 and 2. Therefore using DCM SSR, I am interested in the “most
prominent” average network activity (of a particular known network within
the brain) during the different sessions and how this network activity changes
between the four sessions within each experiment.
For
example if I am comparing say 10 different models to characterize the
“most prominent” network activity, I would like to know how the
model selection changes between all of the sessions within each experiment; and
then compare those model selection results to make inferences between the two
experiments.
What
would be the best way to set up the analysis? Can I run the same 10 model BMS
for each session and see if the winning model changes across sessions? Or
should I concatenate the sessions into one dataset and analyze how A and B
matrix coupling parameters characterize the changes?
Thank
you for the suggestions and help