Hi Yune
> I'm trying to use SnPM to validate the 2nd level random effect analysis
> using SnPM permutation which I have never tried.
> While I can understand the basic concept (e.g., H0: random re-labling A & B
> conditons wouldn't affect the result), I have difficulty figuring out what
> would be the permutated lable when I perform one-sample t-test at the group
> level using contrast images in SnPM. There is only one condition (con image
> where there is only one value at each voxel) and I wonder how it is
> permutated across multiple subjects. Since I have 8 images that are
> submitted to the one-sample t-test, the SnPM shows the maximum permutation
> is 2^8=256, which makes sense if I would figure out how each con image is
> permutated.
For a one-sample t-test, it's the sign (positive or negative) of the
con* image that is randomly assigned; you're testing the null
hypothesis that the distribution of the con* values has a mean of 0.
For much more detail (and an example of a group analysis, just like
what you are doing) see:
Nichols TE, Holmes AP (2001) Nonparametric permutation tests for
functional neuroimaging: A primer with examples. Human Brain Mapping
15, 1-25.
http://dx.doi.org/10.1002/hbm.1058
Hope this helps,
Jonathan
|