Dear FSL users. - I apologize for posting this again but after reading my message from yesterday I realized it might be worth elaborating a little more about some of the problems I am running into with FEATQuerry.
My experiment has 2 runs each with three conditions (A, B & C) and I am interested in extracting percent signal change for an ROI for the contrasts "A > C" and "B > C". I ran my lower level analysis and set up these contrasts as COPE 1 and COPE 2 for each of the runs. I also ran a higher level analysis where I combined the 2 runs and also have COPE1 and COPE 2.
So far so good...
However, I tried running these two methods in FEATQuery and got different results and now I am unsure about the correct way to approach this. I am particularly concerned with two things:
1. My ROI mask has 300 voxels. When I run FeatQuery with the lower level analysis the # of voxels is reduced to 88 while when I run FeatQuery with the higher level (combined runs) COPE the report shows me all the 300 voxels. When I first ran FEATQuerry with the lower level analysis and saw the reduced number of voxels I thought that this was because it only reported on the active voxels for the contrast. If this is the case, why do I see all the voxels when I run the analysis with the higher level COPEs?
2. I was expecting that the average FEATQuery results from combining the lower level analysis would equal (or come very close to) the FEAtQuery results from the higher level analysis. However, this is not the case. When I average the % signal change from run 1 and run 2 (lower level feats) for COPE 1 it gives me a very different result when compared to the higher level analysis that had previously combined these runs.
I hope this makes sense.and I would appreciate any reply as to how people have run this in the past.
Thanking you in advance,
Victor
|