The face-aw contrast (-a 1 -a 2 -c 1 -c 2) is not doing what you think it is doing. It is implementing 0.5 0.5 -0.5 0.5, so you should divide it by 2 before subtracting car-aw to get the equivalent to fac-aw_v_car-aw

doug




On 6/19/14 3:57 PM, DeCross, Stephanie N. wrote:
Hi, 

I’m having trouble figuring out the correct way to weight my conditions with mkcontrast for a couple of complicated contrasts I’m making.

I have six conditions:
#1 female face approach
#2 male face approach
#3 female face withdraw
#4 male face withdraw
#5 car approach
#6 car withdraw

For a couple of contrasts, I had to manually weight with -wcond. For this first one, everything worked. First I made an all faces (male and female, approaching or withdrawing) v. fixation (face_v_fix) contrast, weighted -a 1 -a 2 -a 3 -a 4. I then made an all cars (approaching or withdrawing) v. fixation (car_v_fix) contrast, weighted -a 5 -a 6. To make my contrast face_v_car, which is a subtraction of the above two contrasts, I used the command: mkcontrast-sess -contrast face_v_car -analysis 5.3loom_objns4runs.lh -ncond 6 -wcond 0.25 0.25 0.25 0.25 -0.5 -0.5. When I extracted the data in an ROI analysis for these three contrasts and a bunch others, I manually checked all ROI values for a given subject. The values for my face_v_fix and car_v_fix are correct, and when I manually subtract those two values, it gives me a value identical to the value that derives from the contrast itself. This is as it should be, as they’re two methods of computing the same thing.

My problem is with a similar trio of contrasts. My contrast for all faces (both male and female) approaching minus all faces (male and female) withdrawing (aka face-aw), which is weighted -a 1 -a 2 -c 3 -c 4 in the mkcontrast command, and my car approaching minus car withdrawing contrast (aka car-aw), which is weighted -a 5 -a 6, are both fine. I’m trying to make a contrast face-aw minus car-aw (aka face-aw_v_car-aw), which essentially involves a double subtraction (faces approaching minus faces withdrawing, MINUS cars approaching vs. cars withdrawing) in which the face-aw part involves 4 conditions because of the two genders of faces but the car-aw part involves only 2. The command I used was: mkcontrast-sess -contrast face-aw_v_car-aw -analysis 5.3loom_objsns4runs.lh -ncond 6 -wcond 0.25 0.25 -0.25 -0.25 -0.5 0.5. When I extracted the ROI data, the values for my face-aw and car-aw contrasts were correct, but when I manually subtract them the value is different from the value that derives from the face-aw_v_car-aw contrast itself, when it should be equivalent. I have no idea why it’s not working.

The only thing I can think that would affect this and mess it up is my weighting in the mkcontrast command. Do you know if I’m using the wrong weighting numbers or missing a flag I need or if there's some way to configure the contrast differently I should be doing? I’m sorry if I provided way more information than I needed to or not nearly enough; I’m very new at all things FreeSurfer and finding that the learning curve can be challenging.

Thanks so much,
Stephanie

Stephanie N. DeCross
Clinical Research Coordinator
Psychiatric Neuroimaging Research Program
Martinos Center for Biomedical Imaging
Massachusetts General Hospital
149 13th Street, Charlestown, MA 02129
Phone: 617-724-3283
Fax: 617-726-4078