Bianca,
> I would like to determine the common activations between 2 groups of
> subjects on 1 contrast. I would like to do this as a random effect
> analysis.
>
> The way I see it, there are 2 ways I can go about this:
>
> Do a fixed effect analysis for all subjects (all subjects in 1 design
> matrix) and then do a conjunction analysis on the same contrast over all
> subjects.
This won't yield a random effects analysis, so you don't want this
first option.
> Do a fixed effect analysis with a separate design matrix for each
> subject, create the contrast for each subject. Then do a RFX conjunction
> analysis over the 2 groups for that 1 contrast. I am assuming that
> since I only take 1 contrast for each subject to the second level
> and since my 2 groups consist of different subjects the sphericity
> assumption should hold.
This is the option you want.
> I think the second option is probably the best, but it also presents me
> with a few questions:
> - Would a 2-sample t-test be good enough for me, since I compare 2
> groups on 1 contrast (in the archives, most advice centers around
> ANOVA and regressions)?
> - If so, how do I set up a 2-sample t-test without constant term?
To get a 2-sample t-test with out the constant you run a Basic Models
'One Way ANOVA', specifying two groups.
You can select 'Nonsphericity' if you want to let the variance be
different between the groups, but you definitely don't want 'Correlated
repeated measures', as subjects are independent (Note that 'Replications'
are over 'Repl').
You'll then be able to make separate Group 1 and Group 2 contrasts
and do a conjunction.
-Tom
-- Thomas Nichols -------------------- Department of Biostatistics
http://www.sph.umich.edu/~nichols University of Michigan
[log in to unmask] 1420 Washington Heights
-------------------------------------- Ann Arbor, MI 48109-2029
|