Hi,
I am trying to classify (with SVMs) between two groups using fMRI data. For both groups, we have two tasks, each acquired in separate runs. 
This is slightly different from your tutorial examples, and I have to organize information input differently. Below I describe what I have tried, and I think option 2) is the best. Can you indicate the best option?

1)	I tried to input the different conditions for each participant as different modalities/runs (G1|participant_1|condition_1,condition_2|files_1,files_2), all the conditions have the same design (nº scans and block duration) but the tasks are different . In this case, when I try to apply the “Specify model|Define classes” step, the software only identifies condition_1, in the conditions box, for each participant, even though I had given different names to the conditions.
A workaround would be to try to concatenate all sessions/conditions, and then could use “Specify model|Define classes”. This approaches enabled me to solve the problem described before, but I’m concerned about the influence of concatenating several session as one, and also how the “Data operations” will act on it.

2)	I also tried to input various versions of the same participant, but each with different conditions: participant_1_Cond1, participant_1_Cond2, … and it the becomes possible to select the subject_ condition that I want to define for each class. 

Additionally, I would like to ask on suggested/advised data operations. I think in this experimental paradigm (considering option2)) the best might be to first  mean center features over subjects and then divide data vectors by their norm.

Do you advise, in our case, to use beta values instead of the fMRI scans to compare between groups?

Thanks,
Rui