 |
 |
Dear Julia, With regard to your first issue, yes, when working with differential images the threshold should be a different one, as you will probably encounter not only values close to zero, but also negative values. To solve this you can forward all the different s(m)p1 files into a one-sample t-test, specify the usual thresholds. The mask image from that model could be used as an explicit mask for the other models (no threshold masking, no implicit masking, only the explicit mask then). With regard to the second issue, it sounds as if there's an issue with mean-centering, something like having separate TIV columns for different groups. A regressor that codes group 1 [1 1 1 ... 0 0 0 ...] would correlate highly with a TIV regressor [1200 1300 1400 ... 0 0 0 ...], you have to mean-center the TIV values then. Best Helmut