Dear experts,
I'm very new to fMRI, and I would like to ask for a clarification regarding RFX
and FFX analysis for a group of subjects using SPM5.
I was told that the standard way of running group FFX analysis was to create a
very large design matrix, with all the sessions for all the subjects, and then
look at effects of interest by "collapsing" across subjects.
However, I was under the impression that the only real difference between FFX
and RFX analysis is the way that variance is partitioned. Thus, the difference
is that in RFX we specify treatment effects as *fixed* and subjects as a
*random* factors, while in FFX we only specify *fixed* factors where the levels
of the fixed effects are assumed to be independent and the levels of the
*random* factors are not.
If this is correct, than a group (2nd level) fixed-effects analysis could be
performed by using the summary statistics from the 1st level and importing them
to a 2nd level factorial design and keeping the "independence" option as YES.
Is that correct?
If so, what is the benefit of performing the FFX analysis using a design matrix
with all the sessions for all the subjects, unless you want to look at
conjunctions between subjects?
Any help would be greatly appreciated!
Yerina
|