 |
 |
On Sat, Jun 23, 2012 at 7:42 AM, David V. Smith <[log in to unmask]>wrote: > Be careful when you have more than one input per subject in your > group-level model. If I understand previous posts on this issue, you can > only make valid inferences on the within-subject factor: > https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;eb928af2.1112 > This is correct. Only the time effects can evaluated. This means any comparisons between time points. Because time*group is a time effect, then it can properly be evaluated. The limitation is that randomise can not properly randomise the time*group contrasts, so you can only use OLS. > > With two time points per subject, one solution would be to subtract the > cope images (e.g., t2-t1) and then feed the difference images into > randomize (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=fsl;bef2e9e.1204). > Aside from ignoring the middle time point (e.g., t3-t1), I'm not sure what > you would do with three time points per subject. > This is correct. Randmise could also test group differences at each time point AND the average of all three time points. In summary, you need several models to do the between-subject testing and one test to evaluate the within-subject effects. > > Cheers, > David > > On Jun 22, 2012, at 10:16 PM, Christine Zakrzewski wrote: > > Group differences are the differences btw each group. Measuring the > difference btw the pts of each separate group makes no sense. I think what > you mean is that you want to model a 2 X 3 anova, 2 for the groups and 3 > for sessions--testing for effect of group, session, and the interaction. > Good luck, Christine > > > Date: Fri, 22 Jun 2012 14:48:07 +0100 > > From: [log in to unmask] > > Subject: [FSL] High-level analysis 40 subjects (2 x 20 subjects) with 3 > scan sessions > > To: [log in to unmask] > > > > Dear all, > > > > I am setting up a new analysis and I am trying to figure out the design > of my GLM. > > > > The situation is as follows: I have 2 groups with 20 people in each > group and where each subject has got 3 different scan sessions (fMRI). > > > > So; > > Group 1: 20 subjects with 3 scans for each subject (one week in between > each scan) > > Group 2: 20 subjects with 3 scans for each subject (one week in between > each scan) > > > > I assume this is a three level analysis with 40 subjects each having > three sessions and where we are also interested in comparing both groups? > So the Multi-session/Multi-subject repeated measures? > > > > Basically, I am interested in the differences WITHIN each patient (over > time), the group differences or the differences BETWEEN the patients of > each separate group (mean group 1, mean group 2) and the differences > BETWEEN both groups. > > > > I hope my question is clear.. > > Can anyone help me please? > > > > Best wishes, > > > > Lies > > > -------------------------------------------- > David V. Smith, Ph.D. > Center for Cognitive Neuroscience > Duke University > [log in to unmask] > -------------------------------------------- > > >