Hi,
I'm trying to do a ppi analysis on multi-session data. I was only able to
find one
post pertaining to this in the archives:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind04&L=spm&P=R421090&I=-1
where it is recommended that the sessions should be modelled as one. That
e-mail also
suggests that one can model the individual sessions separately, estimating
one contrast
per session, then forwarding all those contrast to the second level.
I had a different idea, and I wanted to check the legitimacy - I was
thinking of extracting
the PPI regressors separately for each session, then setting up the second
design matrix
to query the interaction itself as a multi-session study. Is there anything
wrong with
this approach? Then, I would only have a single contrast per subject to
forward to the
second level analysis. Also, would it be more appropriate to take a t-test
(with a 1 in
the interaction column of each session, i.e. for a 4 session study the
contrast would be
[1 0 0 1 0 0 1 0 0 1 0 0], or would it be better to use an f-test. (for the
same 4
session study [1 0 0 0 0 0 0 0 0 0 0 0; 0 0 0 1 0 0 0 0 0 0 0 0; 0 0 0 0 0 0
1 0 0 0 0 0;
0 0 0 0 0 0 0 0 0 1 0 0].? Please forgive me if this isnt' clear, I'll be
happy to
clarify.
Thanks so much,
Allison
|