Dear Lisbeth,
N=5 is a very small number for RFX group analysis, and you may run into problems estimating the smoothness of your residuals or the error variance. This would explain why the con image looks fine, but the stats image does not. Have a look at the bottom of your results table. Estimated FWHM and Resel count should be reasonably close to what you would expect after preprocessing. If the voxels are outside the brain it would also help to include an explicit GM+WM mask in your model specification and rerun model estimation.
Hope this helps,
Volkmar
|