Hi Marco, Randomise can be used, but the number of possible permutations would be too small: just 15. However, if the errors are guaranteed to be symmetrical, it's possible to also do sign-flipping. We have an experimental tool that can do it, please, have a look at PALM: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/PALM All the best, Anderson On 13 May 2015 at 22:28, Marco Loggia <[log in to unmask]> wrote: > Hi all, > > I have a PET image from a patient, and would like to compare it voxel-wise > to a group of healthy controls (let's say N=15) to see if the patient is an > 'outlier' in any voxels. > > Ideally, for each voxel, I would perform the nonparametric-equivalent of a > one-sample t-test, with the patients' value at that voxel as the reference > value against which all the controls values are tested. From what I read, > randomize wouldn't seem to do this, but I wonder if I could somehow > implement a strategy similar to that used by the outlier detection > algorithm. > > Thanks! > Marco >