 |
 |
Hi Fatima, if you want to run an ANOVA at the second level (and not a 2 sample t-test as Don suggested), you will have to revert to the PET models in SPM2. Choose PET as the modality and then you will find a greater variety under 'PET'. You will need 'Multi-Group: conditions and covariates'. I think that the entries are pretty straight forward. However, when you choose "Replications are over subjects" the SPM" will configure variance components between groups. These you will have to remove. I have used the attached script in a similar context before and it has worked for me. No guarantees that it'll work for you, but you can try it. Be sure to check you variance components very thoroughly before you estimate the design. Good Luck, Jan Don Rojas wrote: > Fatima, > > Someone with a better stats background than I might serve you better, > but I believe that for the design you mention, with only 2 levels of the > repeated measure (i.e., 1 df), there is no need to do a correction for > non-sphericity, since that assumption is that the correlations between > all possible pairs of samples are identical. Since you only have 1 > pair, you don't have a problem. You should be able to carry out your > SPM2 analysis by performing the repeated measure analysis at the first > level and then use the contrast images from the first level in a second > level analysis such as the 2-sample t-test. For a 2 x 2 design such as > yours it should be essentially equivalent to a 2 x 2 mixed design ANOVA. > > Don > > -----Original Message----- > From: SPM (Statistical Parametric Mapping) [mailto:[log in to unmask]] > On Behalf Of FATIMA HUSAIN > Sent: Tuesday, May 01, 2007 12:29 PM > To: [log in to unmask] > Subject: [SPM] Mixed ANOVA in SPM2 > > I am trying to do a within-subject (task1, task2) and between-subjects > (patients, controls) ANOVA in SPM2. Because of the mixed design, none of > > the standard options in SPM2 basic models work. The problem is that you > need non-sphericity correction for the within-subject factor and not for > > the between-subject one. The number of patients is also fewer than > controls. > > Has anyone had success doing mixed ANOVAs in SPM2? > I see on the archives several work-arounds were posted. Has anyone > published results using any of these alternative methods? > > I tried using Rik Henson's batch script for mixed ANOVA design (posted > in > 2005, same errors with Thilo Kellermann's script) but ran into errors, > pasted below. > > Any help is most appreciated! > > Fatima Husain > p.s. Thanks to everyone for the helpful book suggestions. > > -------------------------------------------------------------- > >>> batch_mixed_anova > > SPM2: spm_spm (v2.66) 12:36:28 - 01/05/2007 > ======================================================================== > Initialising parameters : ...done > ...initialised > Plane 69/69 , block 3/3 : ...done > Temporal non-sphericity (over voxels) : ...REML estimation > ReML Iteration : 1 > ...2.281850e+02 > > SPM2: spm_spm (v2.66) 12:36:37 - 01/05/2007 > ======================================================================== > Initialising parameters : ...computing > s = > > (1,1) 3.3174 > (2,2) 3.3174 > (3,3) 3.3174 > (4,4) 3.3174 > (5,5) 3.3174 > (6,6) 3.3174 > (7,7) 3.3174 > (8,8) 3.3174 > (9,9) 2.6825 > (10,10) 2.6825 > (11,11) 2.6825 > (12,12) 2.6825 > (13,13) 2.6825 > (14,14) 2.6825 > (15,15) 2.6825 > (16,16) 2.6825 > > ??? Index exceeds matrix dimensions. > > Error in ==> spdiags at 114 > a((len(k)+1):len(k+1),:) = [i i+d(k) B(i+(m>=n)*d(k),k)]; > > Error in ==> spm_spm at 421 > s = spdiags(1./sqrt(diag(s)),0,nScan,nScan); > > Error in ==> spm_spm at 830 > SPM = spm_spm(SPM); > > Error in ==> batch_mixed_anova at 158 > SPM = spm_spm(SPM); > -- Jan Gläscher, Ph.D. Div. Humanities & Social Sciences +1 (626) 395-4976 (office) Caltech, Broad Center, M/C 114-96 +1 (626) 395-2000 (fax) 1200 California Blvd [log in to unmask] Pasadena, CA 91125