>Hello SPMers,
>I would classify myself as neophyte, and as such I have two pretty basic >questions for which I think I have the correct answers, but I would like to be >sure.(searching the net has not yielded me particularly good results, but this >may be due to poor search terms).
>I am running a basic block design with 4 stimulus conditions and a rest >condition. I have coded all 5 conditions in the fMRI design.
>1) when i perform a t contrast on any two conditions (e.g. cond #1 & cond #2), >at any give voxel. SPM takes the convolved estimated response of cond #1 and >subtracts cond #2, generating a new model. the resulting t map is how good a >predictor this new model is at any given voxel? OR does it take the average >signal from cond#1 and subtract it from average signal of cond#2.
That depends on how you set up the contrast; for example, you can test the effects of cond. 1 + cond. 2, cond. 1 - cond. 2, cond. 2 - cond. 1, etc. SPM does not 'take the average'; you get what you ask for, assuming you set up your model correctly and your conditions are nonorthogonal-
>2) if I wanted to do a more complex comparison: (e.g. the combination of cond >#1 & cond#2 compared to rest), my entry in the contrast managed would be... >1,1,0,0,-2 (again assuming i have coded 5 conditions).
First of all, I would not recommend comparing active conditons to rest- lots of reasons- mainly you don't know what is going on during that time period (search the list for further discussion on this topic). Also, you want your contrast weights to sum to zero for each contrast- so in your example, it would be .5, .5, 0, 0, -1.
>Thanks in advance.
Sure!
Julie E. McEntee, M.A., C.C.R.P.
Senior Research Program Coordinator
Department of Psychiatry- Neuroimaging
Johns Hopkins University School of Medicine
600 N. Wolfe St./ Phipps 300
Baltimore, MD 21287
Phone: 410-502-0468
Fax: 410-614-3676
>Regards,
>Ryan
|