 |
 |
As long as you fed in the first level contrasts to this model the output will be the difference. Cheers, Jeanette On Wed, Mar 27, 2013 at 12:00 PM, Anders Hougaard <[log in to unmask]>wrote: > Hi Jeanette, > > > Thank you very much! > > I think I set the analysis up correctly (see the design attached) but > the results does not look like the difference in activation between > the to first-level results (which I guess it should?). > Am I doing something wrong? > > Best, > Anders > > > 2013/3/27 Jeanette Mumford <[log in to unmask]>: > > Hi, > > > > Calculate the paired differences in a within-subject, fixed effects > second > > level model. Design matrix > > 1 0 > > 0 1 > > contrast > > 1 -1 > > > > Repeat for all subjects and you have a single paired difference for each > > subject that you can enter into your group model. The group model will > > include a column of 1s and age (mean centered). > > > > Cheers, > > Jeanette > > > > > > On Wed, Mar 27, 2013 at 10:14 AM, Anders Hougaard <[log in to unmask]> > > wrote: > >> > >> Dear FSL experts, > >> > >> > >> Consider this experiment: > >> > >> A group of subjects is scanned during some kind of stimulation before > >> (condition A) and after (condition B) some kind of intervention. > >> > >> I want to know if there is a difference in the activation caused by > >> the stimulation between condition A and condition B. > >> For fMRI data this is easily set up in a paired design using FLAME. > >> > >> The next thing I want to know if a subject's age has an effect on how > >> much the activation changes from condition A to B. > >> Putting subject age into the design as a covariate would cause the > >> design matrix to become rank deficient and generally would not make > >> sense since we are looking at within-subject differences and all > >> subjects have the same age before and after the intervention. > >> > >> How can I find out about the effect of age, or any other > >> subject-specific parameter such as gender or handedness, in this case? > >> > >> > >> All the best, > >> > >> Anders > > > > >