I have got a simple data set of ten subjects scanned twice. What I
would like to check is if there was common activation between all the
subjects (not interested in changes between sessions). I can see three
ways of doing this:
1. Pull all the 20 contrast images into a one sample t-test. This,
however seems wrong since each subject is measured twice. The
inference should be made about the whole population. There is a reason
why we did not scan one person twenty times after all.
2. Average volumes within subjects and do a one sample t-test on those
averages. I believe this was suggested before in a similar problem by
Donald McLaren in this e-mail
3. Do a mixed effect analysis using Flexible Factorial Design. Two
factors: subjects and sessions (dependent) with only the main effects
(see the attachment). The are two issues with this: 1) if I understand
the e-mail cited above it is not statistically valid 2) after fitting
the model when I try to estimate contrast [1 1] (I am not interested
in between session effects, just the overall activation) I get
"invalid contrast" error.
What do you think is the best way to do this?