Dear Jeong,

Dear SPM gurus I have two groups of PET scans (normal & abnormal).Normal group has 6 subjects and abnormal has 8 subjects. And each subject has two scans for hisactivation and baseline.I would like to see the main effects and interactions among groups and conditions.I thought this model could be "Multi-group conditions & covariates" in PET/SPECT model from SPM manual.But in the middle of doing "Multi-group conditions & covariates" I 've met an error messageas below:=======================================================================Initialising design space : ...computing??? Error using ==> spm_spm
This design has no residuals! (df=0)Error in ==> C:\MATLABR11\work\SPM99\spm_spm_ui.m
On line 1427 ==> spm_spm(VY,xX,xM,F_iX0,xC,xsDes)??? Error while evaluating uicontrol Callback.=======================================================================

This error message indicates that your model is under determined, i.e. that you are trying to model more effects than there are scans. It's a bit like trying to fit a second order polynomial to two points.

Variables were set as below:
   - #group = 2
   - #subj = 6(for group1) and 8(for group2) and each subject's scans were selected 
      as his "activation", "baseline" in order in both groups, conditions input were "a r".
   - global normalization = AnCova

I think the Ancova global normalisaton is your problem. As of SPM99 Ancova normalisation is always "by subject", which means you get one global regressor per subject. Together with your subject and condition regressors they will eat all your degrees of freedom. I would recommend simply switching to proportional scaling, and the estimation should run smoothly.

Good luck Jesper