Hello Uta,
basically the random effects analysis is an analysis of the contrast maps
of your first level analysis, e. g. a t-test of one group against zero, or
a t-test for two groups against each other, or some more complex factorial
model etc.
It is VERY difficult to do a power calculation for fMRI - simply, as the
power of your 2nd random effects analysis will depend on the contrast maps
of your first level analysis. As long as you do not know the contrast
value and variance across subjects of this for each voxel for your
paradigm, it will be difficult to state how many subjects you need to find
a group effect.
This contrast value will depend on the field strength, the paradigm (!),
the postprocessing, the type of effect you want to find ("just" a 1-group
effect; or a difference between two groups be it within the same as
repeated measures analysis, or between two groups ), the type of
correction (whole brain vs. small volume correction), so all in all the
N=12 more seem like an empirical value for "most" paradigms.
best whishes,
Philipp
|